Uncomputable trisection invariants of 4-manifolds

arXiv:2608.27811 · math.GT · Submitted 2026-08-28 · 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: We're looking at the paper: "Uncomputable trisection invariants of 4-manifolds".

Emmy: Here's the gist of the paper. Trisection invariants of 4-manifolds are uncomputable because two specific invariants defined by trisections, namely the trisection genus and the Kirby–Thompson L-invariant, cannot be computed by an algorithm for a given triangulated closed orientable 4-manifold.

Ada: The paper establishes that two invariants derived from smooth 4-manifolds using trisections are uncomputable: the trisection genus and the Kirby–Thompson L-invariant.

Emmy: The trisection genus, denoted as g(X), is defined as the minimal genus of a triple intersection surface Σ among all possible trisections of X.

Ada: Furthermore, the Kirby and Thompson L-invariant (LX) is introduced by minimizing the curve-complex complexity of a trisection diagram over all trisections of X.

Paper summary: Ada: So we're looking at this paper titled "Uncomputable trisection invariants of four-manifolds <ref:2608.27811#pg1,trisection invariants of 4-manifolds>." It claims that two specific invariants derived from trisections, the trisection genus and the Kirby–Thompson L-invariant, are uncomputable. This means there is no algorithm that can take a triangulation of a closed orientable four-manifold and give you one of these quantities <ref:2608.27811#pg1,closed orientable 4-manifold and>.

Emmy: That’s right. The paper establishes this by showing that both the trisection genus, which is the minimal genus of a triple intersection surface among all possible trisections, and the Kirby–Thompson L-invariant, which minimizes the curve-complex complexity of a trisection diagram over all possible trisections, are uncomputable.

Kolya: So if an algorithm existed for these invariants, what would it actually be doing? Would it be classifying these four-manifolds in some way <ref:2608.27811#pg1>?

Ada: Exactly. The paper says the proof relies on relating the Kirby–Thompson L-invariant to whether a manifold is diffeomorphic to a connected sum of specific manifolds. If you can compute L, you can tell if the manifold belongs to that specific family of connected sums.

Emmy: And it goes further, according to Theorem one point two, which deals with the PL multisection genus, there's no algorithm for any dimension n greater than or equal to four to output this PL (n minus one)-section genus given a triangulation of a PL n-manifold X <ref:2608.27811#pg2,of a PL n-manifold X>.

Ada: So we have two main results here: the L-invariant is uncomputable, and the multisection genus is uncomputable in dimensions four and higher. This means these invariants are fundamentally beyond what any algorithm can compute for these manifolds.

Kolya: What does this mean for someone who just looks at four-manifolds <ref:2608.27811#pg1>? It suggests that even with a triangulation, you can't always find out these important geometric properties algorithmically.

Ada: It points to a deeper level of complexity in the structure of these manifolds than we might expect. The proof for the L-invariant shows that if L is greater than zero, then the manifold isn't diffeomorphic to a connected sum of S4, CP2, CP2, S1 times S3 and S2 times S2.

Emmy: And if L is zero, it means the manifold *is* one of those specific connected sums. But determining which one it is can be done using its homology groups and intersection form.

Kolya: So the difficulty isn't just calculating a number, but identifying which class of objects you are looking at based on that number.

Ada: Right. And for the multisection genus in dimension four, they use Gordon’s construction. They show that an algorithm computing this genus would be able to tell which members of a specific family of smooth four-manifolds are diffeomorphic to twelve times the connected sum of S2 times S2 <ref:2608.27811#pg3>.

Emmy: That would contradict the fact that those manifolds are not recognizable by any algorithm, which is what Markov and Gordon showed for homeomorphism. This shows that the undecidability property isn't just theoretical; it applies to these geometric properties in both smooth and PL categories.

Kolya: So the proof for Theorem one point two for dimension four is quite concrete, linking the genus value directly to a known undecidable problem about manifold recognition <ref:2608.27811#pg2>.

Ada: It really shows how deep these invariants are rooted in classical unrecognizability results, like Markov's result about the number of moves in S2 times S2 being unrecognizable Mar. This makes you think about what kind of computational limits exist for geometric topology.

Emmy: And this work connects the smooth and PL categories, which is interesting because sometimes those two categories behave very differently under algorithmic scrutiny. The paper points out that while Tancer's results apply in the topological category, this new improvement applies only in the PL category Tan.

Kolya: I wonder about the practical implications here. If we are trying to use these invariants to distinguish manifolds, this paper says we can't build a universal computational tool for them.

Ada: That’s exactly it. The paper is titled "Uncomputable trisection invariants of four-manifolds" because it proves that two specific invariants defined by trisections, the trisection genus and the Kirby–Thompson L-invariant, cannot be computed by an algorithm for a given triangulated closed orientable four-manifold <ref:2608.27811#pg1,the trisection genus and the Kirby–Thompson L-invariant>.

Emmy: It’s about showing that even with the powerful structure theorems provided by trisections—like Gay and Kirby's result that every smooth four-manifold admits one GK—the resulting invariants aren't computable <ref:2608.27811#pg1>.

Kolya: So, to wrap up this segment, we see that the L-invariant is uncomputable because it relates to manifold classification in a way that implies an algorithm recognizing a specific family of connected sums.

Ada: And the multisection genus is uncomputable because for dimension four and higher, it can be used to distinguish members of Gordon's family, which are known not to be algorithmically recognizable.

Emmy: The implication is that these geometric properties defined by trisections are inherently more complex than what a computer can handle through an algorithm given a triangulation.

Kolya: It seems like this paper highlights the boundary between computable geometry and deep topological structure, especially in higher dimensions where we rely on results like Novikov's work on spheres VKF.

Ada: So for someone who just listens to this, the main point is that these key quantities derived from trisections are fundamentally uncomputable for four-manifolds <ref:2608.27811#pg1>. That's what the paper is about.

Conclusion: Ada: So we've looked at how these two specific quantities derived from trisections—the genus and that Kirby Thompson L-invariant—are fundamentally uncomputable for four-manifolds.

Emmy: It really boils down to the fact that even with all the structure trisections give us, we can't always calculate these invariants using a fixed algorithm.

Kolya: That means if you have a triangulation of a four-manifold right now, there’s no guaranteed way for your computer to tell you what that genus or L-invariant actually is.

Ada: Exactly. The paper is about showing these invariants are uncomputable and tying it back to deep recognition problems in topology.

Emmy: It suggests a limit on what we can expect from algorithmic tools trying to fully classify these higher dimensional spaces based on their trisection structure.

Kolya: So, the real question is, how does this affect our ability to study four-manifolds computationally? Does it mean certain geometric questions are just out of reach for current methods?

Ada: It suggests that for these specific invariants, the answer is no. It's not that we lack tools; it's that the structure itself imposes a computational barrier.

Emmy: The authors, Aribi, Courte, Golla, and Moussard specifically point out how these results connect to Gordon’s construction and Markov’s original work on unrecognizability.

Kolya: That link to those classical results is important because it shows this isn't just a new quirk; it fits into the larger picture of what we already know about algorithmic limits in topology.

Ada: It does. It frames the problem not as a lack of ingenuity, but as a fundamental property of these invariants themselves that resists computation by any method we can devise right now.

Emmy: And this whole discussion leads us to think about where else these kinds of structural limitations show up in other areas of mathematics and physics.

University of Illinois Urbana-Champaign · Universidad de Sevilla · Stanford University

math.GT

Submitted: 2026-08-28

Updated: 2026-11-09

Comments: 6 pages, no figures; v2: incorporates referee's suggestions, including a change in title; to appear in Algebraic and Geometric Topology

Project page: https://marckegel.github.io

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 98/100

The gist: Trisection invariants of 4-manifolds are uncomputable because two specific invariants defined by trisections, namely the trisection genus and the Kirby–Thompson L-invariant…

Key concepts

Trisection Genus (g(X))
This is the minimum genus of a triple intersection surface found among all possible trisections of a 4-manifold X. It measures the complexity associated with how X can be decomposed into three pieces using specific surfaces, and its computation is shown to be undecidable.
Kirby–Thompson L-invariant (LX)
This invariant is defined by minimizing the curve-complex complexity of a trisection diagram for a manifold X. If LX > 0, it implies X cannot be written as a connected sum of specific simple manifolds like S4 or CP2, suggesting an algorithmic barrier to calculating its value.
Unrecognizability
This is the idea that there is no algorithm that can reliably tell if one manifold is diffeomorphic to another based on their triangulations. The paper uses results like Markov's theorem and Gordon's construction to show that computing the trisection genus or L-invariant would lead to an algorithm capable of solving these undecidable recognition problems.

Terminology

Summary

Trisection invariants of 4-manifolds are uncomputable because two specific invariants defined by trisections, namely the trisection genus and the Kirby–Thompson L-invariant, cannot be computed by an algorithm for a given triangulated closed orientable 4-manifold. p. 0

Introduction to Invariants

The paper establishes that two invariants derived from smooth 4-manifolds using trisections are uncomputable: the trisection genus and the Kirby–Thompson L-invariant. The trisection genus, denoted as g(X), is defined as the minimal genus of a triple intersection surface Σ among all possible trisections of X. p. 0 Furthermore, the Kirby and Thompson L-invariant (LX) is introduced by minimizing the curve-complex complexity of a trisection diagram over all trisections of X. p. 0

Uncomputability of the L-invariant

Theorem 1.1 states that there does not exist an algorithm that takes as input a PL triangulation and outputs the Kirby–Thompson L-invariant of X. p. 0 The proof relies on relating LX to whether X is diffeomorphic to a connected sum of specific manifolds, showing that if LX > 0, then X is not diffeomorphic to the connected sum of copies of S4, CP2, CP2, S1xS3, and S2xS2. p. 2 This implies that an algorithm computing L would yield an algorithm recognizing a specific manifold family. p. 2 p. 2

Uncomputability of the PL Multisection Genus

Theorem 1.2 asserts that for any dimension n ≥ 4, there is no algorithm to output the PL (n − 1)-section genus of a closed orientable connected PL n-manifold X given its triangulation. p. 1 For n = 4, the proof uses Gordon’s construction of a family of smooth 4-manifolds with Euler characteristic 26, where an algorithm computing the trisection genus would determine which members are diffeomorphic to a specific manifold, contradicting its defining undecidability property. p. 3

Generalization and Related Results

The results extend beyond dimension 4, showing that the PL (n − 1)-section genus g PL n(X) is uncomputable for all n ≥ 4. Moreover, it is shown that the trisection genus g = k is undecidable for every k ≥ 24 by considering Gordon’s family of manifolds. p. 3 Similarly, the PL (n − 1)-section genus g PL n(X) = k is undecidable for all n ≥ 5 and k ≥ 0. p. 3

Foundational Undecidability Tools

The proofs utilize classical unrecognizability results, such as Markov's result that the number of moves in a specific context is unrecognizable. The proof for Theorem 1. p. 22 for n ≥ 5 relies on Novikov's result regarding the recognizability of the PL n-sphere S n. The paper also references results concerning smoothly unknotted non-orientable surfaces in S4 that are algorithmically unrecognizable. p. 3 p. 3

Acknowledgments and Context

The research was conducted during the 2026 Trisectors Workshop at Western Washington University. The work is supported by various grants, including those from the US National Science Foundation and the Simons Foundation. p. 1 The paper notes that for PL 4-manifolds, every manifold admits a smooth structure which is unique up to diffeomorphism. p. 1 The authors acknowledge their support from several institutions and fellowships.

References

[ACGM] F. B. Aribi, S. Courte, M. Golla, and D. p. 3 Moussard.

[BBCD+] S. Bastl, R. Burke, R. Chatterjee, S Dey, et al. p. 3

[BKR] R I Baykur, R Kirby, and D Ruberman. p. 3 p. 3

[CMRZ] S Courte, D Moussard, Q Ren, and X Zhou. p. 4 p. 4

[FHK] S Friedl, T Hirsch, and M Kegel. p. 2 p. 2

[GK] D Gay and R Kirby. p. 0 p. 0

[Gor1] C M Gordon. p. 2 p. 2

[KT] R Kirby and A Thompson. p. 0 p. 0

[Mar] A Markov. p. 2 p. 2

[Mil] J Milnor. p. 1 p. 1

[Tan] M Tancer. p. 2 p. 2

[VKF] I A Volodin, V E Kuznecov, and A T Fomenko. p. 4 p. 4

[BCTT] R Blair, M Campisi, S A Taylor, and M Tomova. p. 4 p. 4

[CT] M Chu and S Tillmann. p. 3 p. 3

--- Page 1 ---

Trisection invariants of 4-manifolds are uncomputable because two specific invariants defined by trisections, namely the trisection genus and the Kirby–Thompson L-invariant, cannot be computed by an algorithm for a given triangulated closed orientable 4-manifold.

[ACGM] F. B. Aribi, S Courte, M Golla, and D Moussard. p. 3

[BBCD+] S Bastl, R Burke, R Chatterjee, S Dey, et al. p. 3

[BKR] R I Baykur, R Kirby, and D Ruberman.

[CMRZ] S Courte, D Moussard, Q Ren, and X Zhou.

[FHK] S Friedl, T Hirsch, and M Kegel.

[GK] D Gay and R Kirby.

[Gor1] C M Gordon.

[KT] R Kirby and A Thompson.

[Mar] A Markov.

[Mil] J Milnor.

[Tan] M Tancer.

[VKF] I A Volodin, V E Kuznecov, and A T Fomenko.

[BCTT] R Blair, M Campisi, S A Taylor, and M Tomova.

[CT] M Chu and S Tillmann.

--- Page 2 ---

This resolves the second part of Problem 4.116 in the K3 problem list [BKR]. p. 1 The first part of that problem remains open, which asks whether a certain distance in the cut complex associated to a given trisection is computable; here, LX is the minimum of this distance over all trisections of X. p. 1

The proof relies on relating LX to whether X is diffeomorphic to a connected sum of specific manifolds, showing that if LX > 0, then X is not diffeomorphic to the connected sum of copies of S4, CP2, CP2, S1xS3, and S2xS2. This implies that an algorithm computing L would yield an algorithm recognizing a specific manifold family.

The proof for Theorem 1.2 for n ≥ 5 relies on Novikov's result regarding the recognizability of the PL n-sphere S n. The paper also references results concerning smoothly unknotted non-orientable surfaces in S4 that are algorithmically unrecognizable.

--- Page 3 ---

In this section, we prove Theorems 1.1 and 1.2 using classical unrecognizability results for manifolds. p. 2 The starting point is a result of Markov asserting that there exists an integer l such that the number of moves in S squared x S squared is unrecognizable [Mar]. p. 2 Here, we say that a smooth compact 4-manifold X is unrecognizable if there exists no algorithm that takes as input a triangulation of a smooth compact 4-manifold X' and outputs whether or not X' is diffeomorphic to X. Currently, the best result of this form is that of Gordon:

Proof Strategy for Theorems

The authors prove Theorems 1.1 and 1.2 by showing that a hypothetical algorithm for computing the L-invariant or the PL (n − 1)-section genus would contradict Theorem 2. p. 21, a slight refinement of it, or its higher-dimensional analogs. p. 2

Proof of Theorem 1. p. 2 p. 21

By [KT, Theorem 12], LX = 0 if and only if X is diffeomorphic to a connected sum of copies of S4, CP2, CP2, S1xS3, and S2xS2. Thus, if LX > 0, then X is not diffeomorphic to the connected sum of copies of S4, CP2, CP2, S1xS3, and S2xS2. p. 2 On the other hand, if LX = 0, then X belongs to the above class of connected sums, and within this class its diffeomorphism type can be determined from its homology groups and intersection form. p. 2 Since both are algorithmically possible (for the latter, see for example [BBCD+, Section 6]), it follows that an algorithm to compute the L-invariant would yield an algorithm recognizing a specific manifold family. p. 2

Proof of Theorem 1.2

It is straightforward to see that the only PL n-manifold with (n −1)-section genus 0 is the PL n-sphere S n. If n ≥ 5, the PL n-sphere is unrecognizable by Novikov; see [VKF, CL]. p. 2 Thus, it is undecidable whether the PL (n − 1)-section genus is equal to 0 for n ≥ 5. p. 2

Construction of Gordon's Family

For n = 4, the authors use Gordon’s construction in [Gor1]. From an Adian–Rabin family of finite group presentations, Gordon constructs a family of closed smooth 4-manifolds with the following properties:

(a) there is no algorithm deciding which Wp’s are diffeomorphic to 12(S squared x S 2).

(b) every Wp has Euler characteristic 26. p. 2

(c) if Wp is not diffeomorphic to 12(S squared x S 2), then π1(Wp) is nontrivial. p. 3

Contradiction for Theorem 1. p. 32 (n=4)

g(12(S squared x S 3)) = 24 as g is subadditive under connected sums and is bounded below by b2. On the other hand, if π1(Wp) is nontrivial, then the lower bound of Chu and Tillmann gives g (Wp) ≥ χ(Wp)-2+3rk(π1(Wp)) ≥ 27. p. 3 Consequently, an algorithm computing the trisection genus would determine which members of Gordon’s family are diffeomorphic to 12(S squared x S S 3), contradicting the defining undecidability property of that family. p. 3

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. Improve algorithmic verification of 4-manifold invariants by leveraging undecidability results for recognition problems to prove that computing invariants like the Kirby–Thompson L-invariant is uncomputable. This allows AI systems to formally recognize when an algorithm for these quantities is impossible, rather than just attempting computation.

  2. Develop a formal system within AI capable of generating and analyzing finite presentations of smooth manifolds to determine if a given invariant calculation leads to a contradiction with known undecidability results, such as the one related to Gordon's family of 4-manifolds. This enables the AI to prove uncomputability for invariants like the trisection genus based on established topological obstructions.

  3. Create a module that uses the unrecognizability property of certain PL n-spheres for n greater than 5 to formally decide whether a given PL (n - 1)-section genus is zero. This system can definitively state that deciding this quantity is impossible for high dimensions, providing a formal boundary for computational tractability.

Sources

Related papers