---
title: Carrollian Lie Algebroids
url: https://www.emergentmind.com/topics/carrollian-lie-algebroids
type: topic
---

# Carrollian Lie Algebroids

Searching arXiv for the specified paper and closely related Carrollian geometry work.
Carrollian Lie algebroids are Lie algebroids endowed with a degenerate metric whose rank-one kernel is carried by a trivial line subbundle, with the distinctive feature that the corresponding Carrollian direction on the base manifold is defined not by the kernel itself but by its image under the anchor map. This construction was introduced to treat singular Carrollian geometries, especially situations in Carrollian gravity and holography where the Carroll vector field vanishes at isolated points and standard Carrollian geometry therefore fails [2510.03877]. In this framework, the null direction remains regular “upstairs” in the algebroid, while its projection to the manifold may collapse, producing a singular Stefan–Sussmann distribution that fluctuates between rank \(1\) and rank \(0\).

## 1. Motivation: from regular to singular Carrollian geometry

In the usual formulation, a weak Carrollian manifold is a triple
\[
(M,g,\kappa),
\]
where \(g\) is a degenerate metric of pointwise type
\[
\operatorname{diag}(1,1,\dots,1,0),
\]
and \(\kappa\) is a nowhere vanishing complete vector field such that
\[
\ker(g)=\operatorname{Span}\{\kappa\}.
\]
This description presupposes that the Carroll vector field is nonzero everywhere. The motivating problem is that in Carrollian gravity and holography one encounters configurations in which the Carroll vector field vanishes at isolated points. At those points the null direction collapses, and the classical definition cannot encode the geometry because the kernel of \(g\) would need to be globally spanned by a nonvanishing vector field [2510.03877].

The basic shift is to move the Carrollian data from the tangent bundle to a Lie algebroid. The kernel of the degenerate metric is retained as a genuine rank-one subbundle, but singular behaviour is transferred to the anchor map. In this way the singularity is not placed in the metric or in the kernel line itself; it is encoded in the image of that line on the base manifold. A common misconception is therefore to identify singular Carrollian behaviour with a degeneration of the kernel bundle. The construction instead keeps the kernel regular and lets the anchor become singular.

This perspective is compatible with the broader contemporary development of Carrollian structures in gravity. For example, three-dimensional AdS-Carroll, Carroll-Galilei, and extended Carroll-Galilei Chern–Simons gravities arise from Lie algebra expansion of two-dimensional Euclidean algebras [2501.00205]. A plausible implication is that Carrollian Lie algebroids occupy the geometric side of a wider program in which Carrollian gauge algebras and Carrollian differential geometry are being developed in parallel.

## 2. Definition and basic algebraic structure

A Carrollian Lie algebroid is a quintuple
\[
(A,[-,-],\rho,g,L),
\]
where \((A,[-,-],\rho)\) is a Lie algebroid over \(M\), \(g\) is a degenerate metric on \(A\), and \(L\subset A\) is a trivial line subbundle such that
\[
\ker(g)=\operatorname{Sec}(L),
\]
with fibrewise diagonalisation of \(g\) of type
\[
\operatorname{diag}(1,1,\dots,1,0),
\]
the final zero direction being precisely the \(L\)-direction [2510.03877]. The explicit inclusion of \(L\) is essential: the Carrollian structure is not determined by \((A,g)\) alone.

A basic structural result is that \(L\to M\) is a Lie subalgebroid. Since \(L\) has rank \(1\), any two of its sections can be written as \(f_1\sigma\) and \(f_2\sigma\) for a frame \(\sigma\), and the Leibniz rule yields
\[
[f_1\sigma,f_2\sigma]
=
\big(f_1\rho_\sigma(f_2)-f_2\rho_\sigma(f_1)\big)\sigma
\in \operatorname{Sec}(L).
\]
Thus \(\operatorname{Sec}(L)\) is closed under the bracket [2510.03877].

Ordinary weak Carrollian manifolds are recovered as a special case by taking \(A=TM\), \(L=\ker(g)\), and \(\rho=\mathrm{id}\). In that situation,
\[
\mathcal{C}=\rho(L)=L,
\]
so the Carroll distribution is regular and nonsingular. Carrollian Lie algebroids therefore strictly generalize ordinary Carroll manifolds [2510.03877].

## 3. The Carroll distribution and singular foliation

The central new geometric object is the Carroll distribution
\[
\mathcal{C}:=\rho(L)\subset TM.
\]
It replaces the classical null direction. Because \(L\) has constant rank but the anchor may vanish on \(L\) at some points, the rank of \(\mathcal{C}\) need not be constant: where \(\rho|_L\neq 0\), the distribution has rank \(1\); where \(\rho|_L=0\), it has rank \(0\). The distribution can therefore fluctuate between rank \(1\) and rank \(0\), precisely capturing a singular Carroll vector field [2510.03877].

The distribution \(\mathcal{C}\) is generally a singular Stefan–Sussmann distribution. It is involutive or integrable in the Stefan–Sussmann sense, but its rank may jump and the dimensions of its leaves may vary. The resulting singular foliation is called the Carroll foliation, and its leaves are Carroll leaves [2510.03877]. Geometrically, rank-\(1\) points behave like ordinary Carrollian points with a null direction, while rank-\(0\) points are locations where that null direction collapses.

A useful dynamical statement concerns \(L\)-paths. If \(\alpha:I\to A\) is an \(L\)-path, meaning \(\alpha(t)\in L\) and \(\rho(\alpha(t))=\dot\gamma(t)\), then the base path \(\gamma\) remains entirely within a single Carroll leaf [2510.03877]. This is presented as support for the interpretation of \(L\)-paths as generalized worldlines of massive particles constrained to move along the Carroll null direction.

The quotient by the null line isolates the transverse metric directions. If
\[
E:=A/L,
\]
then \(g\) descends to a nondegenerate metric \(g_E\) on \(E\) by
\[
g_E(\bar u,\bar v):=g(u,v),
\]
where \(u,v\) are lifts of \(\bar u,\bar v\). This is well defined because adding sections of \(L\) does not change the value of \(g\), and it is nondegenerate because vectors orthogonal to everything lie in \(L\) and therefore vanish in the quotient [2510.03877].

## 4. Symmetries, stationarity, and compatible connections

For a section \(u\in \operatorname{Sec}(A)\), the Lie derivative of the metric is
\[
(\mathcal{L}_u g)(v,w)=\rho_u(g(v,w)) - g([u,v],w)-g(v,[u,w]).
\]
A section is Killing when \(\mathcal{L}_u g=0\). An important result is that if one nowhere-vanishing section of \(L\) is Killing, then every section of \(L\) is Killing. This motivates the notions of a stationary Carrollian Lie algebroid, in which every section of \(L\) is Killing, and a framed Carrollian Lie algebroid, in which a global frame of \(L\) is chosen [2510.03877].

Connection theory is one of the main technical contributions. A Lie algebroid connection \(\nabla\) is called \(L\)-compatible if
\[
\nabla_u v\in \operatorname{Sec}(L)
\quad\text{for all }u\in \operatorname{Sec}(A),\ v\in \operatorname{Sec}(L),
\]
metric compatible if
\[
\rho_u(g(v,w))=g(\nabla_u v,w)+g(v,\nabla_u w),
\]
and Carrollian if both conditions hold [2510.03877].

If a global frame \(\sigma\) of \(L\) is parallel,
\[
\nabla_u \sigma=0 \quad \forall u,
\]
then \(\nabla\) is automatically \(L\)-compatible, because every section of \(L\) is of the form \(f\sigma\) and
\[
\nabla_u(f\sigma)=\rho_u(f)\sigma\in \operatorname{Sec}(L).
\]
Starting from any Lie algebroid connection \(\nabla^0\), one can enforce this by defining
\[
\nabla_u v := \nabla^0_u v - (\nabla^0_u \sigma)\,\omega(v),
\]
where \(\omega\in \operatorname{Sec}(L^*)\) satisfies \(\omega(\sigma)=1\). This yields an \(L\)-compatible connection [2510.03877].

Metric compatibility can also always be achieved. Given any initial connection \(\nabla^0\), its non-metricity is
\[
(\nabla^0_u g)(v,w)=\rho_u(g(v,w)) - g(\nabla^0_u v,w)-g(v,\nabla^0_u w).
\]
A correction tensor \(\Gamma_A\) can then be added so that the modified connection satisfies the metric compatibility equations. Because \(g\) is degenerate, the resulting system is underdetermined, and solutions always exist [2510.03877]. Conversely, metric compatibility already implies \(L\)-compatibility: if \(w\in \operatorname{Sec}(L)\), then \(g(v,w)=0\) for all \(v\), and compatibility forces
\[
g(v,\nabla_u w)=0\quad \forall v,
\]
hence \(\nabla_u w\in \ker(g)=\operatorname{Sec}(L)\).

Combining these statements yields the theorem that every Carrollian Lie algebroid admits a Carrollian connection [2510.03877]. This is an existence statement analogous to the existence part of Levi–Civita theory, but without uniqueness because the metric is degenerate. The corresponding corollary for ordinary geometry is that every weak Carrollian manifold admits a compatible affine connection, and in the framed case one may choose such a connection so that the Carroll frame is parallel. The paper explicitly remarks that torsion-free compatibility is not guaranteed and may be obstructed in degenerate settings.

## 5. Principal examples

The range of examples shows that the definition is designed to capture both regular and genuinely singular situations.

| Source | Construction | Carroll distribution |
|---|---|---|
| Weak Carrollian manifold | \(A=TM\), \(L=\ker(g)\), \(\rho=\mathrm{id}\) | Regular, \(\mathcal{C}=L\) |
| Action algebroid | \(A=M\times\mathfrak g\) with anchor from a Lie algebra action | Singular if the action vector field has kernel |
| Atiyah algebroid | \(A=TP/G\to M\) from a \(G\)-invariant Carrollian structure on \(P\) | Generally singular |
| Mixed null-spacelike hypersurface | \(A=A_0\oplus L\) under simplifying assumptions | Singular Stefan–Sussmann distribution generated by \(\kappa\) |

A Carrollian tangent algebroid is the nonsingular baseline example. Any weak Carrollian manifold \((M,g,\kappa)\) becomes a framed Carrollian Lie algebroid by taking \(A=TM\), \(L=\ker(g)\), and identity anchor, so \(\mathcal{C}=L\) [2510.03877].

Over a point, a Carrollian Lie algebroid reduces to a Lie algebra \(\mathfrak g\) with a one-dimensional subalgebra \(L\subset\mathfrak g\) and a symmetric bilinear form \(g\) such that \(\ker(g)=L\). In this case the Carroll distribution collapses to the point, and ad-invariant Carrollian Lie algebras appear as special cases [2510.03877].

For an action Lie algebroid \(A=M\times \mathfrak g\) associated with a Lie algebra action \(\mathsf a:\mathfrak g\to \mathrm{Vect}(M)\), the anchor is
\[
\rho_{f\zeta}=f\,\mathsf a_\zeta.
\]
If a degenerate metric is chosen with trivial kernel \(L\), then \(\mathcal{C}=\rho(L)\) is generated by the corresponding action vector field; when the action has kernel at some points, the Carroll distribution becomes singular. This provides a direct source of singular Carrollian Lie algebroids from non-free group actions [2510.03877].

Another class begins with a Riemannian Lie algebroid \(A_0\) and a rank-one Lie algebroid \(L\) built from a vector field \(X\). The direct sum
\[
A=A_0\oplus L
\]
is Carrollian with metric inherited from \(A_0\), and \(\mathcal{C}\) is exactly the distribution generated by \(X\), which may be singular when \(X\) vanishes [2510.03877].

The Atiyah algebroid construction is particularly geometric. Let \(\tau:P\to M\) be a principal \(G\)-bundle with right action \(a:P\times G\to P\), and suppose \(P\) carries a \(G\)-invariant Carrollian structure \((\hat g,\kappa)\) satisfying
\[
\hat g((a_g)_*X,(a_g)_*Y)=\hat g(X,Y),\qquad (a_g)_*\kappa=\kappa.
\]
Then the Atiyah algebroid
\[
A=TP/G \to M
\]
inherits a metric
\[
g(u,v)=\hat g(\iota(u),\iota(v)),
\]
where \(\iota:\operatorname{Sec}(A)\hookrightarrow \mathrm{Vect}(P)\) denotes inclusion of invariant vector fields. Since \(\kappa\) is \(G\)-invariant, it descends to a nowhere vanishing section \(\sigma\in \operatorname{Sec}(A)\) spanning a trivial line subbundle \(L\subset A\), and \(\ker(g)=\operatorname{Sec}(L)\). The crucial point is that \(\rho(\sigma)\) may vanish at base points when \(\sigma_p\) becomes vertical, so \(\mathcal{C}=\rho(L)\) is generally singular [2510.03877]. This example makes the anchor-based philosophy explicit: a regular Carrollian structure upstairs can produce singular Carrollian geometry downstairs.

The mixed null-spacelike hypersurface example is designed to model the coexistence of null and spacelike regions. Let \(\Sigma\subset (M,g_M)\) be a three-dimensional hypersurface in a four-dimensional Lorentzian manifold, decomposed into a null part \(\Sigma_{\mathrm{null}}\) with induced metric of type \(\operatorname{diag}(1,1,0)\) and a spacelike part \(\Sigma_{\mathrm{space}}\) with induced metric of type \(\operatorname{diag}(1,1,1)\). Assuming a characteristic vector field \(\kappa\) that is nonzero on \(\Sigma_{\mathrm{null}}\) and zero on \(\Sigma_{\mathrm{space}}\), the construction extends the null line bundle \(L_{\mathrm{null}}\) to a trivial line bundle \(L\to \Sigma\), extends the screen bundle \(S=L_{\mathrm{null}}^\perp/L_{\mathrm{null}}\) to a rank-two bundle \(A_0\to\Sigma\), and sets
\[
A=A_0\oplus L.
\]
With metric
\[
g(u+\phi,v+\chi)=\hat g(u,v),
\]
one has \(\ker(g)=\operatorname{Sec}(L)\), and the anchor is built so that \(L\) is generated by \(\kappa\). Hence \(\mathcal{C}=\rho(L)\) becomes the singular Stefan–Sussmann distribution generated by \(\kappa\) [2510.03877]. The paper stresses that this is a minimal model and that the physically relevant object is the Carroll distribution itself.

## 6. Related developments and extensions

Carrollian Lie algebroids also delimit a conceptual boundary for later generalizations. A direct extension to noncommutative geometry was formulated through Carrollian \(\rho\)-Lie-Rinehart pairs, which generalize Carrollian Lie algebroids to \(\rho\)-commutative, or almost commutative, geometry [2510.19458]. In that setting, a Carrollian structure is a quadruple
\[
(\mathcal{A},\mathfrak g,\mathcal G,\mathfrak l),
\]
where \(\mathfrak l=\ker(\mathcal G)\) is a free cyclic \(\mathcal A\)-submodule generated by a degree-zero section, and the Carroll distribution becomes
\[
\mathcal C:=\mathsf a(\mathfrak l)\subset \rho\mathrm{Der}(\mathcal A).
\]
The quotient metric on \(\mathfrak g/\mathfrak l\), the Killing theory, and compatibility of Carroll connections with the kernel all have direct analogues there [2510.19458].

Two explicit noncommutative examples were constructed: the extended quantum plane and the noncommutative \(2\)-torus. In both cases the theory produces a null direction together with a nondegenerate transverse direction, paralleling the classical Carrollian split [2510.19458]. This suggests that the anchor-based treatment of Carrollian singularity is robust under passage from smooth manifolds to almost commutative algebraic settings.

Within Carrollian gravity more broadly, algebraic constructions based on Lie algebra expansion produce AdS-Carroll, Carroll-Galilei, extended Carroll-Galilei, and post-Carroll-Newtonian algebras together with corresponding three-dimensional Chern–Simons gravity actions [2501.00205]. Those results are not formulated in algebroid language, but they clarify the wider context in which Carrollian geometry is now used: ultra-relativistic limits, degenerate structures, and systematic extensions beyond the standard tangent-bundle framework. A plausible implication is that Carrollian Lie algebroids supply the differential-geometric counterpart to these enlarged Carrollian symmetry structures.

Taken together, these developments place Carrollian Lie algebroids at the centre of a program aimed at taming singular Carrollian geometries. Their defining move is simple but decisive: retain a rank-one degenerate kernel \(L\) in the algebroid, let the anchor determine its manifestation on the base, and identify the Carrollian direction with
\[
\mathcal C=\rho(L).
\]
From this follow the main structural results: \(L\) is a Lie subalgebroid, \(A/L\) carries a nondegenerate quotient metric, the Carroll distribution integrates as a singular Stefan–Sussmann foliation, and compatible Carrollian connections always exist [2510.03877].

Source: https://www.emergentmind.com/topics/carrollian-lie-algebroids