Serre’s intersection-multiplicity conjecture

The explanation

Draw the line y = 0 and the parabola y = x² on graph paper. They touch at the origin and nowhere else. Put y = 0 into the parabola's equation and you get x² = 0. That root x = 0 counts twice, so the curves meet twice at one point. The line y = x behaves differently. Near the origin it crosses the parabola once, with no touching.

Algebra can count these meetings in any setting. Take all power series in x and y, the formulas that describe what happens very close to the origin. This is an example of a regular local ring: the algebraic model of the neighbourhood of a smooth point. A module is an algebraic object that plays the role of a shape through that point. The line and the parabola each give one. Each shape has a dimension, here 1, while the plane has dimension 2.

Jean-Pierre Serre defined a number χ for two such shapes that meet only at the point. It is an alternating sum: a first term, minus a second, plus a third, and so onSource 1. The first term is a length, the number of independent quantities left after imposing both shapes. For the line and parabola, setting y = 0 and y = x² leaves the series 1 and x, so the length is 2. The later terms are corrections called Tor. Here they are all zero, so χ = 2, matching the double meeting.

Serre conjectured that χ is always at least 1 when the two dimensions add up to the dimension of the ring. This is the positivity conjecture. It is not automatic. The correction terms come with alternating signs and could in principle cancel the first termSource 2.

Serre himself proved it in many cases. His argument works in equal characteristic and in the unramified caseSource 3. Equal characteristic means that numbers behave the same way in the ring as at the point, as with power series over the rational numbers. Mixed characteristic means they differ, as in the p-adic integers, where a prime p acts like a small quantity. Ramified means p is especially small: it lies in the square of the ideal of quantities that vanish at the point. Roberts, and separately Gillet and Soulé, proved that χ is zero when the dimensions add up to too littleSource 4. Gabber proved that χ is never negative in the remaining mixed-characteristic caseSource 5. What was missing was ruling out exactly zeroSource 6. Later work covered special ramified cases. Skalit handled rings built smoothly over a two-dimensional baseSource 7. KC and Soto Levins handled shapes that satisfy a lifting conditionSource 8.

This paper claims to prove the conjecture in full, with no limit on characteristic or ramificationSource 9. It does not require either shape to come from a simpler unramified ringSource 10.

The obstacle is the cancellation in the alternating sum. The paper attacks it with two recent toolsSource 11. One is a way of measuring size called normalized length, built from very large perfectoid rings. Under that measure, certain troublesome correction terms count as zero. The other is algebraic K-theory, which compares the ordinary count with the normalized one. The comparison only works after averaging over a group of symmetriesSource 12. The paper shows each term of a first average is positive using a second averageSource 13. That rules out zero.

An earlier public post by Editan asserts a form of the same resultSource 14. The paper says it is credited to Editan, Copilot and GeminiSource 14. It records that post as a claim, not an established theoremSource 14. It also draws consequences for symbolic powers, a way of measuring how ideals growSource 15.

How sure is this? The openai/math release has no Lean formalization for this result, so no computer has checked the proof. The argument relies on deep recent machinery from perfectoid methods and K-theory. No specialist review of the paper has been published.

What it doesn't show. This does not give a formula for the intersection multiplicity; it only shows the number is at least 1.