Inequivalence of the various notions of completeness for Cartan geometries

arXiv:2606.00354 · math.DG · Submitted 2026-05-29 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Mathematics. The day in mathematics: new theorems, new proofs, and what they open.

Ada: I'm Ada.

Emmy: And I'm Emmy, here with Kolya.

Ada: We're looking at the paper: "Inequivalence of the various notions of completeness for Cartan geometries".

Emmy: Here's the gist of the paper. Inequivalence of the various notions of completeness for Cartan geometries demonstrates that three prevailing notions of completeness for Cartan geometries are not actually equivalent.

Ada: The main results provide recipes for constructing a cornucopia of Cartan geometries that satisfy some completeness assumption without also satisfying a stronger one.

Emmy: The Three Notions of Completeness The paper establishes three notions of completeness for Cartan geometries: development-complete, exponential-complete, and ω-constant metric complete.

Ada: Development-completeness is defined as a geometry being dev-complete if every path has a well-defined antidevelopment.

Paper summary: Ada: We're moving on now to the paper "Inequivalence of the various notions of completeness for Cartan geometries". The authors are focusing on showing that the three prevailing notions of completeness for these geometries—development-complete, exponential-complete, and omega constant metric complete—are not actually equivalent.

Emmy: They set up this by defining development-completeness as a geometry being dev complete if every path has a well-defined antidevelopment. Exponential completeness requires all omega constant vector fields to be complete. And then they show that for reductive Cartan geometries, dev complete implies exp complete, and exp complete implies geodesically complete.

Kolya: So the central claim is that in general, none of these three notions are equivalent across the board. That’s a significant finding because it undermines any hope of treating them as interchangeable tools for studying these geometries.

Ada: It matters because they provide recipes for constructing a cornucopia of Cartan geometries that satisfy some completeness assumption without also satisfying a stronger one. They use deformations of the Cartan connection to break these equivalences between the notions.

Emmy: They also give methods for constructing homogeneous examples that are geodesically complete but not exponential-complete, which is a key result. Theorem six point one states that on each Lie group S of dimension at least two, there exists a homogeneous affine geometry that is geodesically complete but not exponential-complete <ref:2606.00354#pg2>.

Kolya: So this means we can find specific structures where the paths are well-behaved enough to be geodesically complete, but the vector fields themselves aren't complete enough for the exponential condition to hold. That’s a concrete failure case.

Ada: And they generalize a construction from earlier work, specifically extending Clifton’s ideas to provide a larger family of examples of geodesically complete Cartan geometries that are not exponential-complete. Theorem six point five proves that on a reductive model and S with dimension equal to the dimension of the quotient G over H, if you have certain conditions involving a g steady map and a half-fine pair, then the resulting geometry is geodesically complete but not exponential-complete <ref:2606.00354#pg2>.

Emmy: That construction involves setting a specific linear map related to the deformation structure, which makes it very specific and applicable to those homogeneous spaces.

Kolya: It sounds like they are building a family of counterexamples that are robust enough for real application rather than just theoretical curiosities. What kind of structures are they using in these constructions?

Ada: They use quadratic vector fields and half-fine pairs to show that a geometry cannot be exponential-complete. That’s the mechanism driving the failure in those specific examples.

Emmy: So, we're seeing how algebraic structure translates directly into completeness properties in a way that isn't always straightforward or guaranteed by standard theorems.

Kolya: It highlights that even when you have nice algebraic setups, there can still be subtleties related to the metric structure that keep things from being universally equivalent.

Conclusion: Ada: So looking at "Inequivalence of the various notions of completeness for Cartan geometries" by Erickson and McKay, the main message is that these three notions of completeness—development-complete, exponential-complete, and omega constant metric complete—are not equivalent in general.

Emmy: The implication is that we shouldn't assume one when applying these concepts outside of a very limited set of examples. Most hopes for using them as interchangeable tools in the broader setting are in vain because they are often too restrictive to be applicable.

Kolya: In simpler terms, it means you can’t just pick any completeness condition and expect the results to hold across the board for these geometries. You have to be very precise about which property you're dealing with.

Ada: Exactly. The paper gives us concrete examples where we can get one kind of completeness without getting another, and it shows exactly why some proofs that claim equivalence are actually flawed because they rely on assumptions about convergence that don't always work out.

Emmy: So the authors are correcting the folklore on this topic by providing detailed explanations of how specious proofs in the literature break down. They offer recipes for constructing examples in the affine case where you can satisfy some completeness assumption without satisfying a stronger one, which is what they set out to do with this work.

Kolya: It’s about making sure that when people talk about these geometric concepts, they are talking about very specific mathematical properties and not just using them as vague labels.

Ada: That’s the core value of the paper here. It moves it from a general statement to providing actionable methods for constructing counterexamples and explaining why certain arguments fail in a more accessible way.

Emmy: So, overall, this work emphasizes that in this field, you need to be very specific about the completeness condition you are dealing with when applying it to complex geometries.

Kolya: And that’s the practical lesson for anyone trying to apply these ideas—be careful about which definition you're using and be ready for cases where they don't match up neatly.

math.DG

Submitted: 2026-05-29

Updated: 2026-11-10

Comments: 26 pages, 5 figures

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 80/100

The gist: Inequivalence of the various notions of completeness for Cartan geometries demonstrates that three prevailing notions of completeness for Cartan geometries are not actually equivalent.

Key concepts

Development-completeness
A geometry is dev-complete if every path has a well-defined antidevelopment. This relates to how paths can be extended or reversed within the geometric structure, ensuring a certain level of structural completeness regarding paths.
Exponential completeness
This requires that all $\omega$-constant vector fields within the geometry must be complete. It is a stronger condition than development-completeness, and for reductive Cartan geometries, dev-complete implies exp-complete.
$\omega$-constant metric completeness
This is another notion of completeness defined for Cartan geometries. The paper shows that this third notion is generally not equivalent to the first two notions discussed, highlighting the variety in how completeness can be defined.

Terminology

Summary

Inequivalence of the various notions of completeness for Cartan geometries demonstrates that three prevailing notions of completeness for Cartan geometries are not actually equivalent. p. 0

Where the paper says it

ERICKSON AND BENJAMIN MCKAY Abstract. We demonstrate that the three prevailing notions of completeness for Cartan geometries are not actually equivalent and provide methods for constructing many examples in the affine case, modernizing and extending results due to Yeaton Clifton.

The main results provide recipes for constructing a cornucopia of Cartan geometries that satisfy some completeness assumption without also satisfying a stronger one. p. 1
Where the paper says it

We hope that, by translating the content of that paper for a more modern audience, we can also help to rectify this disservice. Our main results provide recipes for constructing a cornucopia of Cartan geometries that satisfy some completeness assumption without also satisfying a stronger one.

The Three Notions of Completeness

The paper establishes three notions of completeness for Cartan geometries: development-complete, exponential-complete, and ω-constant metric complete. Development-completeness is defined as a geometry being dev-complete if every path has a well-defined antidevelopment. p. 5

Where the paper says it

Definition 2. A Cartan geometry (G, ω) of type (G, H) is called development-complete or dev-complete if and only if, for e ∈ G, every γ ∈ Pathe(G) has a well-defined antidevelopment γG ∈ Pathe(G).

Exponential completeness requires all ω-constant vector fields to be complete. p. 6
Where the paper says it

A Cartan geometry (G, ω) of reductive type (G, H) is geodesically complete if and only if all of the ω-constant vector fields ω −1 (X) for which X ∈ m are complete. As with development-completeness and exponential-completeness, an exp-complete Cartan geometry (G, ω) of reductive type (G, H) is necessarily geodesically complete because all geodesics are exponential curves.

For reductive Cartan geometries, dev-complete implies exp-complete, and exp-complete implies geodesically complete. p. 6
Where the paper says it

As with development-completeness and exponential-completeness, an exp-complete Cartan geometry (G, ω) of reductive type (G, H) is necessarily geodesically complete because all geodesics are exponential curves. In short, dev-complete implies exp-complete, and exp-complete implies geodesically complete.

The authors show that in general none of these three notions are equivalent. p. 1
Where the paper says it

Both such preprints independently followed the same faulty reasoning used by Kobayashi when he introduced this equivalence result via Theorems 5. 15 of [5], and in an effort to prevent spreading this erroneous result further, we have written this paper to help correct the folklore: in general, none of these three notions of completeness are equivalent.

Deformations to Break Equivalence

The authors provide methods for constructing examples where one notion of completeness is satisfied without satisfying a stronger one. They use deformations of the Cartan connection to break equivalences between these notions. p. 0

Where the paper says it

We demonstrate that the three prevailing notions of completeness for Cartan geometries are not actually equivalent and provide methods for constructing many examples in the affine case, modernizing and extending results due to Yeaton Clifton. In an effort to rectify the folklore on the topic, we also provide detailed explanations of how specious proofs in the literature that would claim that these types of completeness are equivalent break down.

For instance, Theorem 5. p. 0
Where the paper says it

All compact Riemannian manifolds are complete, and even for the noncompact manifolds, there is no sparsity of examples where the condition holds. Moreover, by the Hopf–Rinow theorem, metric completeness for (torsion-free) Riemannian manifolds happens to be equivalent to geodesic completeness, and since we can convert between the two types of completeness indiscriminately, the assumption gives us plenty of utility for relatively little cost.

7 shows that for every smooth circle embedding γ: S1→R m∼=Aff(m)/GLm R, there exists a deformation ϕ such that the deformed geometry is geodesically complete but not development-complete. p. 1
Where the paper says it

Theorem (Theorem 5. For every smooth circle embedding γ : S 1 ,→ R m ∼= Aff(m)/ GLm R, there exists a (smooth) deformation ϕ of the Klein geometry (Aff(m), ωAff(m) ) of type (Aff(m), GLm R) over

Constructing Homogeneous Examples

The paper constructs examples of homogeneous affine geometries that are geodesically complete but not exp-complete. This is achieved by utilizing quadratic vector fields and half-fine pairs to show that a geometry cannot be exp-complete. p. 6

Where the paper says it

A Cartan geometry (G, ω) of reductive type (G, H) is geodesically complete if and only if all of the ω-constant vector fields ω −1 (X) for which X ∈ m are complete. As with development-completeness and exponential-completeness, an exp-complete Cartan geometry (G, ω) of reductive type (G, H) is necessarily geodesically complete because all geodesics are exponential curves.

Theorem 6.1 states that on each Lie group S of dimension at least 2, there exists a homogeneous affine geometry that is geodesically complete but not exp-complete. p. 2
Where the paper says it

Theorem (Theorem 6. On each Lie group S of dimension at least 2, there exists a homogeneous affine geometry that is geodesically complete but not exponential-complete.

Generalizing the Clifton Plane Construction

The construction for the Clifton plane is generalized to provide a larger family of examples of geodesically complete Cartan geometries that are not exp-complete. Theorem 6.5 proves that on a reductive model and S with dim(S) = dim(G/H), if ρ: s → g is g-steady and there exists a half-fine pair for ρm(ξ) ∈ m of the form (ρh(ξ), Z), then (S × H, ωρ) is geodesically complete but not exp-complete. p. 0

Where the paper says it

First, when a model (G, H) is reductive—meaning there exists an AdH-invariant subspace m ⊆ g complementary to h, so that g = m ⊕ h as an H-representation—we can define a notion of geodesic for Cartan geometries (G, ω) of type (G, H) by taking exponential curves in G of the form t 7→ exp(tω−1 (X))g for X ∈ m and g ∈ G, and then projecting down to the base manifold. In the Riemannian case, this definition is equivalent to the usual notion of geodesic, and just like we do for Riemannian manifolds, we can define a reductive Cartan geometry to be geodesically complete whenever all of these geodesics exist for all time.

This construction involves setting a specific linear map ρh: s → gldim(S)R by defining the matrix structure corresponding to the deformation.

Path Approximation and Faulty Proofs

The paper analyzes the idea of approximating paths through development, showing that faulty proofs hinge on erroneously assuming convergence. Proposition 3. p. 8

Where the paper says it

Of course, this leads us to ask whether a similar procedure works with arbitrary Lie groups G, and hence to the following result. Proposition 3.

2 shows that if (G, ω) is exp-complete, then for each δ > 0, the path γδ admits a well-defined antidevelopment. p. 9
Where the paper says it

If (G, ω) is exp-complete, then for each δ > 0, the path γδ ∈ Pathe(G) defined in Proposition 3. 1 by concatenating exponential paths admits a well-defined antidevelopment (γδ)G ∈ Pathe(G).

The analysis in Section 3 demonstrates that this convergence of approximating antidevelopments is not guaranteed in general. p. 5
Where the paper says it

ERICKSON AND BENJAMIN MCKAY ((γk)G) converge to (γ∞)G in Pathe(G). In Section 3, we will investigate when this implication is reversible, so that convergence of paths in the model geometry can tell us something about convergence of their antidevelopments.

Lemma 3.4 shows that if the images of approximating paths are contained in a compact subset C, then the developments converge to a path with development (γ∞)G. p. 9
Where the paper says it

If we could show that these antideveloped Euler approximation paths (γδ)G ∈ Pathe(G) always converge in Pathe(G), then this would prove that exponential-completeness implies development-completeness, as then we would have a well-defined antidevelopment for every smooth path in Pathe(G), which would give a well-defined antidevelopment for every (piecewise smooth) path in Pathe(G) after reparametrization. Crucially, however, this convergence is not actually guaranteed, even though the derivatives are well-behaved enough that the developments converge.

This suggests that convergence is not always guaranteed, which is where faulty proofs fail. p. 9
Where the paper says it

Crucially, however, this convergence is not actually guaranteed, even though the derivatives are well-behaved enough that the developments converge. This is where the faulty proofs of this supposed equivalence fall apart: they erroneously assume that this convergence necessarily happens.

Where the Gains Come From

The core of the paper lies in rectifying folklore by providing concrete constructions and detailed explanations of why specious proofs break down. The authors provide recipes for constructing geometries that satisfy some completeness assumption without satisfying a stronger one. p. 1

Where the paper says it

We hope that, by translating the content of that paper for a more modern audience, we can also help to rectify this disservice. Our main results provide recipes for constructing a cornucopia of Cartan geometries that satisfy some completeness assumption without also satisfying a stronger one.

By translating the content of Clifton's work for a more modern audience, they correct the disservice done to the topic. The results offer recipes for constructing examples in the affine case. p. 0
Where the paper says it

ERICKSON AND BENJAMIN MCKAY Abstract. We demonstrate that the three prevailing notions of completeness for Cartan geometries are not actually equivalent and provide methods for constructing many examples in the affine case, modernizing and extending results due to Yeaton Clifton.

References

[1] Andreas Cap and Jan Slovák, Parabolic Geometries I: Background and General Theory, Mathematical Surveys and Monographs, vol. 154, Amer. Math. Soc., Providence, RI, 2009.

[2] Yeaton H. p. 0

Where the paper says it

ERICKSON AND BENJAMIN MCKAY Abstract. We demonstrate that the three prevailing notions of completeness for Cartan geometries are not actually equivalent and provide methods for constructing many examples in the affine case, modernizing and extending results due to Yeaton Clifton.

Clifton, On the completeness of Cartan connections, Journal of Mathematics and Mechanics 16 (1966), no. p. 0
Where the paper says it

ERICKSON AND BENJAMIN MCKAY Abstract. We demonstrate that the three prevailing notions of completeness for Cartan geometries are not actually equivalent and provide methods for constructing many examples in the affine case, modernizing and extending results due to Yeaton Clifton.

6, 569–576.

[3] Jacob W. p. 0

Where the paper says it

INEQUIVALENCE OF THE VARIOUS NOTIONS OF COMPLETENESS FOR CARTAN GEOMETRIES JACOB W. ERICKSON AND BENJAMIN MCKAY Abstract.

Erickson, Holonomy of parabolic geometries near isolated higher-order fixed points, 2025. p. 0
Where the paper says it

INEQUIVALENCE OF THE VARIOUS NOTIONS OF COMPLETENESS FOR CARTAN GEOMETRIES JACOB W. ERICKSON AND BENJAMIN MCKAY Abstract.

[4] Joachim Hilgert and Karl-Hermann Neeb, Structure and Geometry of Lie Groups, Springer Monographs in Mathematics, Springer, New York, NY, 2012. p. 8

Where the paper says it

In other words, Euler approximation works by approximating the solution of an ODE by a concatenation of exponential paths for the Klein geometry (R m, ωRm ) of type (R m, {0}). Of course, this leads us to ask whether a similar procedure works with arbitrary Lie groups G, and hence to the following result.

[5] Shoshichi Kobayashi, Theory of connections, Annali di Matematica Pura ed Applicata 43 (1957), 119–194. p. 1

Where the paper says it

Indeed, this supposed equivalence seems widely accepted in the subject’s folklore, and both of the authors of this paper have previously written preprints that included attempts to prove this result. Both such preprints independently followed the same faulty reasoning used by Kobayashi when he introduced this equivalence result via Theorems 5.

[6] Benjamin McKay, Complete complex parabolic geometries, Int. p. 0

Where the paper says it

INEQUIVALENCE OF THE VARIOUS NOTIONS OF COMPLETENESS FOR CARTAN GEOMETRIES JACOB W. ERICKSON AND BENJAMIN MCKAY Abstract.

Math. Res. Not. (2006).

[7] An Introduction to Cartan Geometries, 2024. p. 0

Where the paper says it

In an effort to rectify the folklore on the topic, we also provide detailed explanations of how specious proofs in the literature that would claim that these types of completeness are equivalent break down. Introduction In Riemannian geometry, completeness is a comfortable and fruitful assumption.

[8] R. W. Sharpe, Differential Geometry: Cartan’s Generalization of Klein’s Erlangen Program, Graduate Texts in Mathematics, vol. p. 3

Where the paper says it

Similarly, we can model manifolds with a choice of affine connection (M, ∇) on affine geometry (Aff(m), GLm R), where Aff(m) ≃ R m ⋊ GLm R is the group of affine transformations of R m ∼= Aff(m)/ GLm R. The geometric structure of a model geometry (G, H) can be encoded as a Cartan geometry called the Klein geometry.

166, Springer-Verlag, New York, NY, 1997.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. System can construct examples of homogeneous affine geometries that are geodesically complete but not exp-complete by utilizing quadratic vector fields and half-fine pairs. This allows for a systematic search for geometries lacking development completeness while preserving geodesic completeness.

  2. The system can identify when a Cartan geometry (G, omega) is not development complete by checking if the antidevelopment of a path is not guaranteed to exist, or by testing if an exponential curve is not well-defined for all time.

  3. The system can use deformations described in Lemma 5.2 and Theorem 5.6 to systematically construct new examples that break completeness notions along specific curves, such as showing that (Aff(m), omegaAff(m) + phi) is not development complete when deformed along a circle embedding of R/Z.

  4. The system can apply the Trotter product formula approximation methods, specifically Proposition 3.3 and Lemma 3.4, to check if the convergence of approximating developments is guaranteed under specific conditions like g-steadiness or path compactness.

  5. The system can analyze the relationship between completeness notions by identifying that dev-complete implies exp-complete, and exp-complete implies geodesically complete for reductive types, helping to refine search strategies for non-equivalent completeness properties.

Abstract

We demonstrate that the three standard notions of completeness for Cartan geometries are not actually equivalent, modernizing and extending results due to Yeaton Clifton. In an effort to rectify the folklore on the topic, we also provide detailed explanations of how specious proofs in the literature that would claim that these types of completeness are equivalent break down. In the affine case, we construct geodesically complete homogeneous examples with an inextensible exponential curve for each connected Lie group of dimension at least 2, as well as an infinite-dimensional family of examples whose exponential curves are defined for all time but whose overlying frame bundle is not complete as a metric space with respect to a natural metric for the Cartan connection.

Sources

Related papers