Recursos
Resumen
Take a Hodge-generic curve in the Siegel threefold and explain why it can meet the quaternionic special curves only finitely many times, with no condition on its boundary. You also see why the height bound has to be polynomial in the degree for the argument to work.
The problem and the points that count
A₂ has dimension three. A curve and a special curve should meet in expected dimension 1 + 1 − 3 = −1, so Zilber–Pink predicts finitely many intersections, even when the special curves vary [0:08]. A QM point is one whose full geometric endomorphism algebra equals an indefinite quaternion division algebra over ℚ. The split algebra M₂(ℚ) does not count, and neither does a quaternion algebra that only embeds in a larger algebra [1:02]. Earlier results needed a boundary stratum, a large-Galois-orbit assumption, or bounded supersingular proximity. This proof needs none of these, because its height bound is proved, not assumed [1:35].
The engine in three moves
- Rank-five variation [2:30]: trace-zero self-adjoint operators with signature (3,2) and Hodge numbers (1,3,1). The period lies on a 3-dimensional quadric, and at a QM point it lies on the conic cut out by the positive plane Eₜ.
- Height bound [3:15]: h(t) ≤ c₁[K(t):K]^{c₂}, obtained from auxiliary-function interpolation around an interior center t₀.
- Counting [4:01]: d conjugates give at least d/J points in one chart, with heights at most T = c₃d^B. Above c(η)T^η points, Habegger–Pila gives a definable path.
- Think-first [4:53]: the count wins for any η < 1/B. With an exponential bound, no η works.
- Contradiction [5:23]: along the path, the conics sweep out a set of dimension at most 2. Full monodromy (André) makes the period image Zariski dense in dimension 3. So degrees are bounded, and Northcott finishes [5:54].
To use on your next unlikely-intersection proof: ask where the polynomial height bound comes from, what path the count produces, and which monodromy fact rules that path out.