Zilber-Pink Conjecture on Unlikely Intersections

Zilber-Pink Conjecture on Unlikely Intersections

OpenAI's claim

The selected manuscripts address Hodge-generic curves in the Siegel threefold $\mathcal A_2$ over $\overline{\mathbb Q}$. Manuscript 1 claims the full finiteness assertion for intersections with special subvarieties of dimension at most one. Manuscripts 2 and 3 give the quaternion-division and elliptic-product-with-CM branches, respectively. These are curve cases of the Shimura intersection question; manuscript numbers follow Sources.

Mathematical statement of the claimed result

Manuscript 1.

Let $\mathcal A_2$ be the moduli space of principally polarized abelian surfaces over $\overline{\mathbb Q}$, and let $C$ be any nonempty reduced irreducible closed Hodge-generic algebraic curve therein. The complete asserted conclusion is \[ \#\left(C(\overline{\mathbb Q})\cap \bigcup_{\substack{Z\subset\mathcal A_2\ \mathrm{special}\\\dim Z\le1}} Z(\overline{\mathbb Q})\right)<\infty. \] In particular the points whose surface is isogenous to an elliptic square are finite. Two additional assertions give finiteness for surfaces isogenous to $E_1\times E_2$ with a CM factor, and for surfaces having a full rational endomorphism algebra that is quaternionic, a division algebra and indefinite. There are no bounds on CM fields, isogeny degrees, endomorphism orders or polarizations, and no height, boundary or integrality restrictions.

Manuscript 2.

Let $C$ be a closed algebraic curve, reduced, irreducible and Hodge generic, in the Siegel threefold $\mathcal A_2$ over $\overline{\mathbb Q}$, the asserted finiteness is \[ \#\{s\in C(\overline{\mathbb Q}): \operatorname{End}_{\overline{\mathbb Q}}(A_s)\otimes\mathbb Q \text{ is an indefinite quaternion division algebra over }\mathbb Q\}<\infty. \] The full endomorphism algebra, not merely a contained quaternion order, is specified; the quaternion algebra and order may vary.

Manuscript 3.

Take a closed curve $C$ over $\overline{\mathbb Q}$, with its reduced structure and assumed irreducible, in $\mathcal A_2$, the moduli of principally polarized abelian surfaces. Assume Hodge genericity: $C$ lies in no proper special subvariety. The claimed finite set is \[ \{s\in C(\overline{\mathbb Q}):A_s\text{ is isogenous to }E_1\times E_2, \text{ with at least one }E_i\text{ having CM}\}. \] Both the CM fields and the isogeny degrees may vary; the theorem does not restrict to a fixed product locus.

Outline of the argument

For manuscript 1 (family 016; 2026-09-24), two auxiliary markings control bad-center proximity in a five-dimensional orthogonal variation. Common-parameter interpolation gives arithmetic height bounds, and orbit counting handles elliptic squares; classification of special curves then joins this branch to the CM-product and quaternionic branches.

For manuscript 2 (family 016; 2026-09-24), a rank-five variation detects the endomorphism locus. Two-argument interpolation and an ordinary-point zero estimate produce a height estimate unaffected by how many places exhibit supersingular closeness; Galois-orbit counting then gives finiteness.

For manuscript 3 (family 016; 2026-09-24), canonical product isogenies and degeneration comparisons bound heights along the curve. Period estimates produce a Galois-orbit lower bound, which conflicts with the definable counting bound unless only finitely many such points occur.

Formal verification

No Lean formalization is listed at this snapshot for manuscripts 1, 2 or 3 (family 016; 2026-09-24).

This site has not executed the Lean code. Described statement coverage is not an independent proof check.

Status caveat

The manuscripts were produced by an internal OpenAI model. OpenAI cautions that results without formalization can contain mistakes. This site has not independently checked the proofs; the label records the scope of OpenAI's claim.

Sources

• OpenAI, Elliptic squares and Zilber–Pink for curves in A ${}_{2}$, 2026-09-24; citation key OAI:Elliptic-Squares-and-Zilber-Pink-for-Curves-in-A2-September-24-2026. PDF (GitHub); PDF (direct); Manuscript page and citation; Latest version; Statement source: intro:squares; Statement source: intro:zp.
• OpenAI, Quaternionic division points on curves in the Siegel threefold, 2026-09-24; citation key OAI:Quaternionic-Division-Points-on-Curves-in-the-Siegel-Threefold-September-24-2026. PDF (GitHub); PDF (direct); Manuscript page and citation; Latest version; Statement source: thm:main.
• OpenAI, The E×CM component of Zilber–Pink for curves in A ${}_{2}$, 2026-09-24; citation key OAI:The-E-times-CM-Component-of-Zilber-Pink-for-Curves-in-A2-September-24-2026. PDF (GitHub); PDF (direct); Manuscript page and citation; Latest version; Statement source: thm:main.
• OpenAI catalogue entry.

Zilber-Pink Conjecture on Unlikely Intersections

Definitions and mathematical statement

A special subvariety of a mixed Shimura variety is one arising from the appropriate Shimura subdatum and Hecke-translate construction. For irreducible $W\subseteq S$, the expected intersection dimension with $T$ is $\dim W+\dim T-\dim S$. The record calls an irreducible $A\subseteq W\cap T$ atypical if \[ \dim A>\dim W+\dim T-\dim S. \] The claim is that only finitely many such $A$ are maximal under inclusion. Equivalently their union is the union of finitely many atypical subvarieties. The source's special-subvariety convention is essential; replacing "special" with an arbitrary algebraic family would give a different problem.

Formulation and concept references: [S1].

Short English statement

Are all unexpectedly large special intersections inside a fixed Shimura subvariety controlled by finitely many maximal exceptional pieces?

Research attempt

Process, scope, and first gap.

This attempt was produced by GPT-6 Astra Ultra on 27 September 2026, by checking the cited formulation, reducing the dimension inequality, and testing an attempted finiteness argument against explicit algebraic examples. The calculations below are elementary reductions and applications of named theorems, with no novelty claim. Write $s=\dim S$ and $w=\dim W$. Subvarieties are closed and irreducible unless otherwise indicated. Work in the relevant irreducible component of $S$.

The first unproved assertion is already concrete for a curve $W$ contained in no proper special subvariety of $S$: \[ \#\left(W\cap\bigcup_{\substack{T\text{ special in }S\\ \operatorname{codim}_S T\ge2}}T\right)<\infty. \tag{82.1} \] For arbitrary mixed Shimura $S$, this attempt does not prove $82.1$. In higher dimension one must also control maximal positive-dimensional atypical pieces. The objective is finiteness for each fixed $W$, with no asserted uniform bound across all $W$.

Literature and the arithmetic input.

The ambient-dimension convention is that of [S1, Conjecture 2.3]. The stronger hypotheses in the non-$\overline{\mathbb Q}$ results of [S2] cannot be discarded to obtain an assertion for every algebraic $W$. One established input used below is Pila's André--Oort theorem for products of modular curves [E1, Theorem 1.1*]: the irreducible components of the Zariski closure of a set of special points are special. The theorem applies to subvarieties defined over $\mathbb C$; its special points and special subvarieties have their modular meaning. Only the modular-surface consequence is used here. No general mixed-Shimura theorem is inferred from that case.

Replacing arbitrary pieces by components.

Suppose $A\subset W\cap T$ satisfies the strict inequality in the statement. Choose an irreducible component $B$ of $W\cap T$ containing $A$. Then \[ \dim B\ge\dim A>w+\dim T-s, \] so $B$ is atypical too. Consequently every maximal atypical $A$ is an actual component of some intersection. Moreover every atypical $A$ is contained in a maximal atypical subvariety: among atypical subvarieties containing $A$, choose one of largest dimension. A proper containment between irreducible closed subvarieties strictly increases dimension, so this choice is maximal.

This gives the equivalence of the two finiteness formulations without taking the Zariski closure of the exceptional union. If finitely many atypical $A_i$ cover that union, an irreducible maximal atypical $B$ is contained in one $A_i$, because an irreducible space cannot be a finite union of proper closed subsets. Maximality then gives $B=A_i$. Conversely finitely many maximal pieces cover all atypical pieces by the preceding dimension argument. The bounded dimension proves existence of maximal elements; it does not bound how many there are.

For any fixed finite list $T_1,\ldots,T_m$, the assertion holds immediately. Each $W\cap T_i$ has finitely many irreducible components. Every atypical piece witnessed by this list lies in one of the finitely many atypical components. Thus the missing issue is uniform control as the special subvariety varies through the entire infinite collection.

Two boundary reductions and a codimension test.

If $W$ is contained in a proper special $T_0$, then \[ w>w+\dim T_0-s. \] Hence $W$ itself is atypical and is the unique maximal atypical piece. This case is genuinely trivial for the recorded ambient convention. It would be a mistake to silently replace $S$ by $T_0$ and claim the resulting, usually stronger, question had been solved. If $W=S$, on the other hand, any $A\subset T$ has $\dim A\le\dim T$, and there are no atypical pieces.

For a proper $A\subset W$, set $r=w-\dim A$. The atypicality test becomes the exact integer inequality \[ \operatorname{codim}_S T>r. \tag{82.2} \] Thus an atypical divisor of $W$ must be witnessed by a special subvariety of ambient codimension at least two. An atypical point requires ambient codimension at least $w+1$. In particular, intersecting a curve with a special divisor at a point has the expected dimension zero and is not atypical merely because the point is arithmetically interesting. The strict inequality is essential.

When $w=0$, there is at most one maximal piece, namely $W$ itself if it is atypical. When $s=1$, irreducible $W$ has dimension zero or is the whole component, so these boundary arguments settle that ambient dimension completely.

Exact reduction for curves.

Assume $w=1$ and that $W$ is contained in no proper special subvariety. A positive-dimensional irreducible closed $A\subset W$ equals $W$. It cannot be atypical: a witness $T$ would have to contain $W$ and be proper in $S$. Every atypical piece is therefore a point.

For a point $a\in W\cap T$, its atypicality is \[ 0>1+\dim T-s, \] which is precisely $\operatorname{codim}_S T\ge2$. Conversely every point in $82.1$ satisfies this inequality. Distinct points do not contain one another, so each is maximal. This proves that $82.1$ is exactly the original finiteness assertion in this case, rather than just a necessary condition. It also identifies why a proof that controls only positive-dimensional intersections leaves the first difficult case untouched.

A completed modular-surface subcase.

Take $S=Y(1)^2$, the product of two coarse modular curves, and an arbitrary irreducible algebraic $W\subset S$. If $W$ has dimension zero or two, use the boundary reductions. If $W$ is a curve contained in a proper special subvariety, $W$ itself is atypical as above.

In the remaining curve case, a special $T$ of codimension at least two is a special point: a pair of complex-multiplication $j$-invariants. Were infinitely many such points to lie on $W$, they would be Zariski dense in the irreducible curve. The theorem from [E1] would then make $W$ special, contrary to the assumption. Therefore the set is finite, and the full recorded assertion holds in this modular surface. The elementary work is the reduction to the exact theorem hypothesis; the finiteness theorem is an external input, not established anew here.

The same argument does not extend by simply writing $Y(1)^n$ in place of $Y(1)^2$. For $n\ge3$, codimension-two special varieties can have positive dimension. In $Y(1)^3$, for example, \[ Y(1)\times\{c_2\}\times\{c_3\} \] is special when $c_2,c_3$ are CM values. Its points with non-CM first coordinate are not special points of the product. A point of intersection with a curve nevertheless has the strict dimension excess required by $82.2$. A special-point theorem therefore does not cover all the intersections appearing in the curve reduction.

A failed Noetherian argument, with a countertest.

A tempting proposal is to enumerate special $T_i$, form the finite union of atypical components obtained up to step $m$, and appeal to Noetherianity to conclude that this increasing sequence of closed sets stabilizes. Noetherianity gives stabilization for decreasing chains of closed sets, not increasing ones.

The failure can be made algebraic and dimensionally exact. Let $S=\mathbb A^2_{\mathbb C}$, $W=\{y=0\}$, and choose the family \[ T_n=\{(n,0)\},\qquad n\in\mathbb Z_{\ge1}. \] Relative to this deliberately arbitrary family, every $T_n$ is an atypical point of $W$, since $0>1+0-2=-1$. The union of the first $m$ points is closed and strictly increases with $m$. Their infinite union is dense in $W$, because a nonzero univariate polynomial has only finitely many roots, but is not closed. None of the individual witnesses contains $W$; the closure $W$ is not an atypical piece for this family.

This is not a counterexample to Zilber--Pink: these $T_n$ are not being asserted to be Shimura-special. It isolates the exact invalid step in an argument that uses only finite-dimensional algebraic geometry. The same example shows that bounding the degree and number of components of each separate intersection would not suffice: here each intersection has degree one and a single component.

Verification and next obstruction.

The dimension calculations were checked at codimensions zero, one, and two and at $w=0,1,s$. Empty intersections contribute no irreducible pieces. The component argument never takes the closure of the infinite exceptional union. The modular result uses the theorem only where every remaining witness is zero-dimensional; it does not identify all atypical points with special points in higher dimension. These are symbolic checks, not a computation over CM values or an exhaustive verification of Shimura subvarieties.

A useful next target is an arithmetic complexity bound on the codimension-at-least-two witnesses meeting a fixed curve, strong enough to imply $82.1$. A second target is a finite collection of proper atypical loci containing every positive-dimensional exceptional component for a fixed higher-dimensional $W$. Neither target follows from degree bounds on one intersection or from bounded lengths of inclusion chains. A potential counterexample would need infinitely many incomparable maximal pieces satisfying the actual special-subvariety conditions, rather than the arbitrary points in the countertest.

Outcome.

UNRESOLVED. The component and codimension reductions are proved, the modular-surface case follows from the stated André--Oort input, and the proposed purely Noetherian route is refuted. The finiteness assertion for a general mixed Shimura variety remains unproved.

Sources

• S1 J. Pila, 'Combining the conjectures of Schanuel and Zilber-Pink', Rend. Lincei Mat. Appl. 36 $2025$, 653-670, DOI 10.4171/RLM/1086; accepted 1 July 2025. https://ems.press/content/serial-article-files/52570. Formulation source.
• S2 The Zilber-Pink conjecture for varieties not defined over Qbar. https://arxiv.org/html/2504.00865v1. Further reference; not a proof endorsement.
• S3 A note on unlikely intersections in Shimura varieties. https://arxiv.org/html/2209.07967. Further reference; not a proof endorsement.
• E1 J. Pila, O-minimality and the André--Oort conjecture for $\mathbb C^n$, Annals of Mathematics 173 $2011$, 1779--1840, Theorems 1.1 and 1.1*. https://annals.math.princeton.edu/wp-content/uploads/annals-v173-n3-p11-p.pdf. Established input for products of modular curves; consulted 27 September 2026.

Zilber-Pink Conjecture on Unlikely Intersections

Definitions and mathematical statement

A special subvariety of a mixed Shimura variety is one arising from the appropriate Shimura subdatum and Hecke-translate construction. For irreducible $W\subseteq S$, the expected intersection dimension with $T$ is $\dim W+\dim T-\dim S$. The record calls an irreducible $A\subseteq W\cap T$ atypical if \[ \dim A>\dim W+\dim T-\dim S. \] The claim is that only finitely many such $A$ are maximal under inclusion. Equivalently their union is the union of finitely many atypical subvarieties. The source's special-subvariety convention is essential; replacing "special" with an arbitrary algebraic family would give a different problem.

Formulation and concept references: [S1].

Short English statement

Are all unexpectedly large special intersections inside a fixed Shimura subvariety controlled by finitely many maximal exceptional pieces?

Research attempt

Current frontier and special-subvariety convention.

Pila's Section 2.3 was inspected for the mixed Shimura formulation and its countable collection of special subvarieties [E1]. The full complex-field statement must not be replaced by a number-field case. The work on varieties not defined over $\overline{\mathbb{Q}}$ in [S2] addresses that distinction, but its particular reduction and hypotheses cannot simply be declared a theorem for every mixed Shimura variety. The counting ingredient used below is the Pila--Wilkie bound on the transcendental part of a fixed definable set [E2].

Attack route.

A tempting qualitative route says that the atypical union is algebraic by Noetherianity. This is false without a closedness theorem. An arithmetic route compares large collections of conjugates with subpolynomial definable point counts, but needs bounded-complexity coding and control of semialgebraic blocks. The attempt proves a precise closedness lemma and an explicit counting criterion, then identifies the missing geometric bridges for the exact all-complex, mixed setting.

Attempt: closedness would be enough, but is not automatic.

Write $U\subseteq W$ for the union of all atypical subvarieties. There are only countably many special $T$. Each intersection $W\cap T$ has finitely many irreducible components. If an irreducible $A\subseteq W\cap T$ is atypical, an irreducible component $B$ of that intersection containing it satisfies $\dim B\geq\dim A$ and the same strict dimension inequality, hence is itself atypical. Therefore $U$ is a countable union $\bigcup_i B_i$ of closed irreducible atypical subvarieties. The countability here is for special, not arbitrary weakly special, subvarieties [E1].

Lemma. For such a countable family over ${\mathbb{C}}$, if $U$ is Zariski closed, it is a finite union of atypical subvarieties. To prove it, write the closed set $U$ as a finite union of its irreducible components $Z_1,\ldots,Z_s$. An irreducible complex variety cannot be a countable union of proper closed subsets. One proof takes an affine open, a finite Noether normalization onto affine space, and images of the proposed proper closed subsets. These images are proper closed sets of smaller dimension. All their defining coefficients lie in one countable subfield of ${\mathbb{C}}$. A point of affine space whose coordinates are algebraically independent over that field avoids all of them; such coordinates can be chosen successively because each algebraic closure of a countable field is countable. A point in the finite nonempty fibre of the normalization then avoids the proposed cover. Dimension zero is immediate.

Apply this fact to $Z_j=\bigcup_i(Z_j\cap B_i)$. Some $B_i$ contains $Z_j$. Since $B_i\subseteq U$ and $Z_j$ is a maximal irreducible component of $U$, $B_i=Z_j$. Thus every $Z_j$ is atypical, proving the lemma. The converse, that a finite union is closed, is immediate. This supplies an exact closedness reformulation in this setting.

The missing statement is precisely that $U$ is closed. Noetherianity controls descending chains of closed sets and finite irreducible decompositions of a closed set, not arbitrary unions. An infinite countable set of complex points in ${\mathbb{A}}^1$ is already a counterexample to the naive inference: every point is closed, but their union is not closed. Taking the closure of $U$ can add subvarieties not yet shown atypical, which is the issue the conjecture requires one to solve.

Attempt: a finite-complexity counting criterion.

Consider a collection of candidate maximal atypical objects, possibly subvarieties encoded by parameters, with complexity $\Delta\geq1$. Suppose one can prove all the following for this fixed $W$: there is a single definable set $Z\subseteq{\mathbb{R}}^m$; every candidate yields at least $c\Delta^\delta$ distinct rational codes in $Z^{\rm tr}$; their heights are at most $C\Delta^b$; and the relation from codes to candidates has finite fibres. Here $b,\delta,c,C>0$ are fixed, and $Z^{\rm tr}$ excludes the positive-dimensional connected semialgebraic subsets. The Pila--Wilkie theorem states \[ \#\{z\in Z^{\rm tr}\cap{\mathbb{Q}}^m:H(z)\leq H\}\ll_{Z,\varepsilon}H^\varepsilon \qquad(\varepsilon>0) \] [E2]. Therefore each candidate would satisfy \[ c\Delta^\delta\leq C_{Z,\varepsilon}(C\Delta^b)^\varepsilon. \] Choose $\varepsilon=\delta/(2b)$. It follows that $\Delta$ is bounded uniformly over the candidates. The height bound then places every candidate above at least one member of a finite set of rational codes. Finite fibres make the candidate collection finite.

This is a rigorous conditional counting argument, including a finiteness condition sometimes lost after obtaining bounded complexity. A bound on complexity alone is not useful if infinitely many candidates may share that complexity and the same code. Rational codes are an explicitly sufficient convention here, not an assertion that the actual mixed Shimura objects already possess such codes. Versions with bounded-degree algebraic codes need their corresponding counting theorem and degree bounds; neither is silently substituted.

Attempt: a completely checked toric calibration.

As a separate toric analogue, let $W\subset({\mathbb{C}}^\times)^2$ be $x+y=1$ and let special subvarieties mean torsion cosets. The curve is not contained in a proper torsion coset: an identity $x^a(1-x)^b=\zeta$ on $W$ would have zero orders at $x=0$ and $x=1$, forcing $a=b=0$. Thus positive-dimensional intersections with proper torsion cosets are not possible. The atypical points in this two-dimensional ambient torus are its torsion points on $W$. If both $x$ and $y=1-x$ are roots of unity, $|x|=|1-x|=1$, and hence \[ 1=|1-x|^2=2-2\operatorname{Re}(x). \] So $\operatorname{Re}(x)=1/2$ and $x=(1\pm i\sqrt3)/2$, with $y$ its conjugate. Both possibilities really are roots of unity. There are exactly two such points. The calculation illustrates the need for extra arithmetic rigidity to make a countable special-point union finite; countability alone did not do it. It is not offered as a solution for an arbitrary mixed Shimura variety.

Stress test.

A lower bound for Galois orbits cannot be used without showing that the relevant conjugates stay in the fixed $W$. For arbitrary $W/{\mathbb{C}}$, that stability is not automatic. Spreading out, field-of-definition parameters, and special closures must be handled rather than replacing $W$ by a variety over $\overline{\mathbb{Q}}$. The source's scope explicitly retains these complex varieties.

Also, definable counting permits many points on its semialgebraic part. If the hypothetical codes land there, the displayed upper bound does not apply. One would need a functional-transcendence or block-analysis theorem turning such families into larger atypical subvarieties and then exploit maximality. Positive-dimensional maximal subvarieties need their own parameter spaces, degree control and finite-fibre coding; a theorem only about isolated special points does not settle their finiteness. None of those geometric bridges is supplied by the elementary counting inequality.

Outcome.

Outcome: Conditional reduction.

The attempt proves the closedness-to-finiteness lemma under the exact countability convention and a quantitative coding criterion that would force finitely many maximal objects. It checks a small toric analogue and identifies why Noetherianity and raw Galois-orbit language do not close the general argument. No new mixed Shimura case is proved. A completion needs the actual closedness theorem or the missing coding, orbit, and block control for every $W/{\mathbb{C}}$ in the original scope; the elementary reductions carry no priority claim.

Sources

• S1 J. Pila, 'Combining the conjectures of Schanuel and Zilber-Pink', Rend. Lincei Mat. Appl. 36 $2025$, 653-670, DOI 10.4171/RLM/1086; accepted 1 July 2025. https://ems.press/content/serial-article-files/52570. Catalogue formulation source.
• S2 The Zilber-Pink conjecture for varieties not defined over Qbar. https://arxiv.org/html/2504.00865v1. Catalogue research/status reference; not a proof endorsement.
• S3 A note on unlikely intersections in Shimura varieties. https://arxiv.org/html/2209.07967. Catalogue research/status reference; not a proof endorsement.
• E1 Jonathan Pila, Combining the conjectures of Schanuel and Zilber--Pink, Rendiconti Lincei -- Matematica e Applicazioni 36 $2025$, 653--670, Section 2.3 and the discussion of special subvarieties. https://ems.press/content/serial-article-files/52570. Exact mixed-Shimura formulation inspected 27 September 2026; the toric example in this notebook is explicitly only an analogue.
• E2 Jonathan Pila and Alex J. Wilkie, The rational points of a definable set, Duke Mathematical Journal 133 $2006$, 591--616, main counting theorem. https://eprints.maths.manchester.ac.uk/942/. Primary institutional manuscript record inspected 27 September 2026; the theorem is used only on a hypothetical fixed definable transcendental part with rational codes and explicit height bounds.