A counterexamples to the Polishchuk--Van den Bergh conjecture on curves

arXiv:2610.10147 · math.AG, math.CT · Submitted 2026-10-07 · 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: "A counterexamples to the Polishchuk--Van den Bergh conjecture on curves".

Emmy: Here's the gist of the paper. For every h ≥ 1, an explicit S3 action on the smooth projective model of y squared = t(12h−5 + t) has quotient of genus h and exactly one ramification orbit, with inertia of order three.

Ada: This construction provides counterexamples to Conjecture A of Polishchuk and Van den Bergh by showing that the conclusion fails for effective actions on smooth projective curves.

Emmy: Counterexamples to the Conjecture The paper constructs an explicit S3 action on a smooth projective model of y squared = t(12h−5 + t) where the genus of the curve is 6h − 3 and the genus of its quotient is h.

Ada: This action exhibits exactly one nonfree orbit with a stabilizer of order three, and its fixed quotient for the three-cycle class consists of two points, but this category admits no C-linear exact fully faithful embedding into the equivariant derived category.

Paper summary: Ada: To summarize what this paper does, they build these explicit examples for every h greater than or equal to one. They take the smooth projective model of y squared equals t raised to the power of twelve h minus five plus t, and apply an S3 action.

Emmy: The key feature they highlight is that this action results in a quotient curve whose genus is exactly h, and there's only one ramification orbit involved in this setup, which has an inertia order of three.

Kolya: So it’s not just any curve; it’s tied to this specific power of t and S3 symmetry. What's the main claim they are pushing against?

Ada: The paper claims that for every h greater than or equal to one, there is an explicit S3 action on this curve model that has exactly one ramification orbit with inertia order three.

Emmy: And the result of this construction is a quotient curve of genus h, and the authors show that its derived category simply admits no C-linear exact fully faithful embedding into the equivariant derived category.

Kolya: So they’re saying that even when all the setup conditions for their conjecture are met, this specific type of structure—this one with inertia order three—is impossible in terms of derived categories.

Ada: Exactly. They show that this failure happens for every h greater than or equal to one, and they prove more generally that the maximum number of pairwise completely orthogonal exceptional objects is P i floor(ei over two), where ei are the inertia orders.

Emmy: This is really pointing towards a structural limitation in how these derived categories can interact with these group actions on curves. It suggests a constraint on what kind of collections we can have.

Kolya: And they also characterized when a conjugacy-class decomposition with embeddings linear over the coarse quotient exists, stating it only happens if the number of elements from the conjugacy class that hit every branch inertia group is less than or equal to floor(ei over two).

Conclusion: Ada: So, looking at this paper, "A counterexamples to the Polishchuk--Van den Bergh conjecture on curves," it’s clear that this work tackles a deep question about the structure of derived categories associated with group actions on curves.

Emmy: The authors are using these specific S3 actions on curves with inertia order three to show that Conjecture A of Polishchuk and Van den Bergh is false. They achieve this by finding a specific curve, like the genus three curve y squared equals t raised to the power of seven plus t, whose quotient is an elliptic curve b squared equals u cubed minus four u.

Kolya: For someone who just listens to the show, what does this mean in practice? It’s not just abstract math about derived categories; it means there are specific geometric objects where these expected structural properties don't hold.

Ada: Right. The implication is that the conditions for a linear action decomposition might be more restrictive than previously thought, or that the simple assumption of smoothness and effectivity isn't enough to guarantee the conclusion of their conjecture.

Emmy: They also point out that these counterexamples don't apply to linear reflection-group actions, which is a bit nuanced. They show this failure occurs specifically in the context of non-linear actions, like the S3 action they built.

Kolya: So what’s the takeaway for how we think about these kinds of geometric objects? It suggests we have to be much more careful when we assume certain properties about fixed quotients and their derived categories.

Ada: That's right. This paper gives us a concrete family of counterexamples, proving that the conclusion of Conjecture A doesn't hold in those cases, even when the setup looks perfect on the surface.

Beijing Jiaotong University

math.AG, math.CT

Submitted: 2026-10-07

Updated: 2026-11-04

Comments: All comments are welcome!

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

Importance score: 82/100

The gist: For every h ≥ 1, an explicit S3 action on the smooth projective model of y squared = t(12h−5 + t) has quotient of genus h and exactly one ramification orbit, with inertia of order three.

Key concepts

S3 action
This refers to an explicit action of the symmetric group on three elements (S3) on a geometric object. The paper uses this specific type of action to construct curves where the conjecture's conclusion does not hold, specifically showing that certain embeddings into derived categories are impossible.
Conjecture A
This is a conjecture by Polishchuk and Van den Bergh concerning the existence of certain types of embeddings (C-linear exact fully faithful embeddings) related to group actions on smooth projective curves. The paper demonstrates that their conclusion does not always follow, even under the stated hypotheses.
Inertia order
This describes a property related to how an element in the inertia group acts on the curve's structure during a covering or ramification process. The paper focuses on actions with exactly one ramification orbit and an inertia of order three, which is crucial for their counterexample construction.
Perf(Y)-linear embeddings
These are specific types of embeddings between derived categories related to the geometry of the curve Y. Theorem 5.3 provides a condition under which these embeddings can be realized, linking them to inequalities involving genus and inertia groups.

Terminology

Summary

For every h ≥ 1, an explicit S3 action on the smooth projective model of y squared = t(12h−5 + t) has quotient of genus h and exactly one ramification orbit, with inertia of order three. p. 0

Where the paper says it

We construct effective finite group actions on smooth projective complex curves for which Conjecture A of Polishchuk and Van den Bergh fails. For every h ≥ 1, an explicit S3 action on the smooth projective model of y 2 = t 12h−5 + t has quotient of genus h and exactly one ramification orbit, with inertia of order three.

This construction provides counterexamples to Conjecture A of Polishchuk and Van den Bergh by showing that the conclusion fails for effective actions on smooth projective curves. p. 0
Where the paper says it

COUNTEREXAMPLES TO THE POLISHCHUK–VAN DEN BERGH CONJECTURE ON CURVES SHENGYONG PAN Abstract. We construct effective finite group actions on smooth projective complex curves for which Conjecture A of Polishchuk and Van den Bergh fails.

Counterexamples to the Conjecture

The paper constructs an explicit S3 action on a smooth projective model of y squared = t(12h−5 + t) where the genus of the curve is 6h − 3 and the genus of its quotient is h. This action exhibits exactly one nonfree orbit with a stabilizer of order three, and its fixed quotient for the three-cycle class consists of two points, but this category admits no C-linear exact fully faithful embedding into the equivariant derived category. p. 0

Where the paper says it

For every h ≥ 1, an explicit S3 action on the smooth projective model of y 2 = t 12h−5 + t has quotient of genus h and exactly one ramification orbit, with inertia of order three. The fixed quotient for the three-cycle class consists of two points, but its derived category admits no C-linear exact fully faithful embedding into the equivariant derived category.

Consequently, all the smoothness and effectivity hypotheses of Conjecture 1. p. 1
Where the paper says it

2 SHENGYONG PAN (4) there is no C-linear exact fully faithful functor Db (Spec C ⨿ Spec C) −→ Db S3 (Ch). Consequently, all the smoothness and effectivity hypotheses of Conjecture 1.

1 hold, but its conclusion fails. p. 1
Where the paper says it

Consequently, all the smoothness and effectivity hypotheses of Conjecture 1. 1 hold, but its conclusion fails.

Bounds on Exceptional Objects

Theorem 5.3 proves that Conjecture 1. p. 1

Where the paper says it

Let Hi be an inertia group at that orbit and ei = |Hi |. 3 proves that Conjecture 1.

1 admits a realization by Perf(Y)-linear embeddings if and only if [g] ∩ Hi ≤ j ei/2k for every i and [g] ≠ [1]. p. 1
Where the paper says it

3 proves that Conjecture 1. 1 admits a realization by Perf(Y )-linear embeddings if and only if |[g] ∩ Hi | ≤ j ei 2 k for every i and [g] ̸= [1].

This criterion has no genus assumption and applies to smooth connected quasiprojective complex curves. p. 1
Where the paper says it

1 admits a realization by Perf(Y )-linear embeddings if and only if |[g] ∩ Hi | ≤ j ei 2 k for every i and [g] ̸= [1]. This criterion has no genus assumption and applies to smooth connected quasiprojective complex curves.

Lemma 2.2 establishes the maximum number of pairwise completely orthogonal exceptional objects supported on an orbit of inertia order e as floor(e/2). p. 3
Where the paper says it

(4) Lemma 2. Among the exceptional objects supported on D, the maximum number that can be pairwise completely orthogonal is ⌊e/2⌋.

Decomposition Structure

The ordered curve decomposition is known to consist of the coarse-curve category together with ei − 1 exceptional objects at each branch orbit. The decomposition DbG(C) = Dπ∗Db(Y), B1,. p. 0

Where the paper says it

More generally, when the coarse quotient has positive genus, we prove that the maximum number of pairwise completely orthogonal exceptional objects is P i ⌊ei/2⌋, where ei are the inertia orders. We also characterize conjugacy-class decompositions with embeddings linear over the coarse quotient: such a decomposition exists if and only if |[g] ∩ Hi| ≤ ⌊ei/2⌋ for every nonidentity conjugacy class and every branch inertia group Hi.

.., BbE, Bi = ⟨Mi,ei−1,.., Mi,1⟩ is established. p. 6
Where the paper says it

Each exceptional subcategory here is admissible: Hom to and from its generator has finitedimensional total cohomology because the generator has proper finite support. The exceptional projection triangles, together with the established orthogonality and generation, yield the semiorthogonal decomposition (7).

The ramification subcategory is Db(mod CA2), which contains an ordered exceptional pair where the two generators are not completely orthogonal. p. 6
Where the paper says it

The ramification subcategory is Db (mod CA2). It contains an ordered exceptional pair, whereas its two generators are not completely orthogonal.

Criterion for Linear Embeddings

Theorem 5.3 provides the necessary and sufficient condition for a conjugacy-class decomposition with embeddings linear over the coarse quotient to exist, which is [g] ∩ Hi ≤ j ei/2k for every branch orbit Di. p. 0

Where the paper says it

More generally, when the coarse quotient has positive genus, we prove that the maximum number of pairwise completely orthogonal exceptional objects is P i ⌊ei/2⌋, where ei are the inertia orders. We also characterize conjugacy-class decompositions with embeddings linear over the coarse quotient: such a decomposition exists if and only if |[g] ∩ Hi| ≤ ⌊ei/2⌋ for every nonidentity conjugacy class and every branch inertia group Hi.

When these conditions hold, all embeddings can be given by G-equivariant kernels on the natural fibre products (Cg/CG(g)) ×Y C. p. 8
Where the paper says it

(11) When these conditions hold, the identity block may be chosen as π ∗Db (Y ), and the nonidentity blocks may occur in any prescribed total order. All embeddings can be given by G-equivariant kernels on the natural fibre products (C g /CG(g)) ×Y C.

This implies that for a linear action on a vector space V, every fixed quotient is irreducible, so the original and connected-component formulations coincide. p. 7
Where the paper says it

For a linear action on a vector space V , the variety V g/CG(g) is irreducible, being a finite quotient of a vector space. Thus the original and connected-component formulations coincide in that setting.

Connected Component Formulation

Conjecture 4.2 states that there is a total order on I(X, G) and a semiorthogonal decomposition of DbG(X) with subcategories A([g],Z) ≃ Db(Z), one for each pair in I(X, G). p. 7

Where the paper says it

2 (Connected-component formulation). There is a total order on I(X, G) and a semiorthogonal decomposition of Db G(X) with subcategories A([g],Z) ≃ Db (Z), one for each pair in I(X, G).

Theorem 4.3 confirms that Conjecture 4. p. 0
Where the paper says it

COUNTEREXAMPLES TO THE POLISHCHUK–VAN DEN BERGH CONJECTURE ON CURVES SHENGYONG PAN Abstract. We construct effective finite group actions on smooth projective complex curves for which Conjecture A of Polishchuk and Van den Bergh fails.

2 holds for every effective finite group action on a smooth connected quasiprojective complex curve C. p. 7
Where the paper says it

Conjecture 4. 2 holds for every effective finite group action on a smooth connected quasiprojective complex curve C.

Categorical Grouping

Theorem 5.3 states that the following conditions are equivalent: (1) Conjecture 1. p. 8

Where the paper says it

Let a finite group G act effectively on a smooth connected quasiprojective complex curve C. The following conditions are equivalent: (1) Conjecture 1.

1 admits a realization whose embeddings are Perf(Y)-linear for the natural maps of the fixed quotients to Y = C/G, and (2) For every branch inertia group Hi and every nonidentity conjugacy class, [g] ∩ Hi ≤ j ei/2k. p. 8
Where the paper says it

1 admits a realization whose embeddings are Perf(Y )-linear for the natural maps of the fixed quotients to Y = C/G. (2) For every branch inertia group Hi and every nonidentity conjugacy class, |[g] ∩ Hi | ≤ j ei 2 k .

When these conditions hold, the identity block may be chosen as π∗Db(Y), and the nonidentity blocks may occur in any prescribed total order. p. 8
Where the paper says it

(2) For every branch inertia group Hi and every nonidentity conjugacy class, |[g] ∩ Hi | ≤ j ei 2 k . (11) When these conditions hold, the identity block may be chosen as π ∗Db (Y ), and the nonidentity blocks may occur in any prescribed total order.

This allows for a grouping of local decompositions into a global semiorthogonal decomposition. p. 9
Where the paper says it

□ Thus the local inequalities characterize the existence of a conjugacy-class decomposition with linear embeddings. The construction allows the exceptional generators within each branch block to change; it does not require grouping the original objects Mi,l without modification.

Linearity and Conclusion

Corollary 5.4 states that if [g] ∩ Hi ≤ 1 for every branch inertia group Hi and every nonidentity conjugacy class [g], then Conjecture 1. p. 9

Where the paper says it

Suppose that |[g] ∩ Hi | ≤ 1 for every branch inertia group Hi and every nonidentity conjugacy class [g]. Then Conjecture 1.

1 holds, with Perf(Y)-linear embeddings given by kernels on the natural fibre products. p. 9
Where the paper says it

Then Conjecture 1. 1 holds, with Perf(Y )-linear embeddings given by kernels on the natural fibre products.

This applies to finite abelian groups acting effectively on smooth connected quasiprojective complex curves. p. 9
Where the paper says it

1 holds, with Perf(Y )-linear embeddings given by kernels on the natural fibre products. In particular, this applies to finite abelian groups acting effectively on smooth connected quasiprojective complex curves.

The local inequalities characterize the existence of a conjugacy-class decomposition with linear embeddings. p. 9
Where the paper says it

For an abelian group, every conjugacy class is a singleton. □ Thus the local inequalities characterize the existence of a conjugacy-class decomposition with linear embeddings.

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. Improved AI can detect failures of Conjecture A by constructing effective finite group actions on smooth projective complex curves for which Conjecture A fails, specifically by finding cases where the fixed quotient for the three-cycle class consists of two points, but its derived category admits no C-linear exact fully faithful embedding into the equivariant derived category.

  2. Improved AI can calculate the maximum number of pairwise completely orthogonal exceptional objects in a given equivariant derived category using Theorem 1.4, which states that ocDbG(C) = X b i=1 j ei/2k when genus(Y) >= 1, and this bound is attained by the simples with indices 0, 2,..., 2⌊e/2⌋ − 2.

  3. Improved AI can characterize conjugacy-class decompositions with embeddings linear over the coarse quotient by checking if [g]∩Hi ≤ j ei/2k for every nonidentity conjugacy class and every branch inertia group Hi, as stated in Theorem 5.3, which is equivalent to Conjecture 1.1 admitting a realization whose embeddings are Perf(Y)-linear for the natural maps of the fixed quotients to Y = C/G.

  4. Improved AI can implement a connected-component formulation for semiorthogonal decompositions by applying Conjecture 4.2, which states that There is a total order on I(X, G) and a semiorthogonal decomposition of DbG(X) with subcategories A([g],Z) ≃ Db(Z), one for each pair in I(X, G).

  5. Improved AI can verify the sufficiency of the connected-component formulation by checking if the identity fixed quotient is the connected curve Y and if every connected component of a nonidentity fixed quotient is a point, matching them with the exceptional objects Mi,ei−1,..., Mi,1.

  6. Improved AI can determine when Conjecture 1.1 holds for finite abelian groups by using Corollary 5.4, which states that if [g]∩Hi ≤ 1 for every branch inertia group Hi and every nonidentity conjugacy class [g], then Conjecture 1.1 holds, with Perf(Y)-linear embeddings given by kernels on the natural fibre products.

Abstract

We construct an effective S 3 action on the smooth projective model of y 2=t 12h-5+t for every h at least1. Its quotient has genus h, and the fixed quotient for the three-cycle class consists of two points. We prove that the derived category of these two points admits no-linear exact fully faithful embedding into the equivariant derived category. This gives a counterexample to Conjecture A of Polishchuk and Van den Bergh. More generally, for a positive-genus quotient we compute the maximum number of completely orthogonal exceptional objects as i e i/2, where e i are the inertia orders. We also give a necessary and sufficient condition for conjugacy-class decompositions with embeddings linear over the quotient, and construct kernels on the natural fibre products. For genus-zero quotients we obtain Euler-form bounds and exact capacities for all-two ramification with at least four branch points and for types (2,3,3), (2,3,4) and (2,3,5).

Sources

Related papers