First-hand information for everyone
Main discovery A proof shows that for r ≥ 10 and very general points p1,…,pr in the complex projective plane P2, any nonzero homogeneous polynomial of degree d that vanishes to order at least m ≥ 1 at each pi must satisfy d > m √r. (This is Nagata’s conjecture as stated.) (F1, F10)
How the proof works (high level) The argument first reduces the interpolation problem to a local analytic model. It uses the standard reduction to the normal cone and a collision of r fat points into a single weighted monomial scheme. This replaces the original question about forms on P2 by an interpolation question on the normal ruled surface of a smooth plane cubic. (A fat point is a point with prescribed multiplicity; a monomial scheme is defined by a monomial ideal.) (F2, F5)
The proof then studies the asymptotics of theta functions as the smooth cubic degenerates to a nodal cubic. Those asymptotics, together with an interpolation duality, reduce the local analytic interpolation problem further to a simple combinatorial interpolation problem on a triangle of lattice points. The theta-function analysis produces local bases that converge to exponentials e^{k w + j y}, and the resulting restriction matrices converge to an explicit combinatorial matrix. (F3, F7, F6)
A concrete interpolation statement (Proposition 2.4) is used at the combinatorial stage. Under the numerical hypotheses q > 9 p and K ≥ 3 h √(pq), interpolation on the triangle holds: a certain restriction matrix has full rank, so arbitrary values on the triangle can be interpolated by the indicated polynomials. This combinatorial fact is a key step in the chain of reductions. (F4)
Transferring back to the plane The analytic construction on the normal ruled surface yields a vanishing statement: for suitable local holomorphic coordinates (w,y) on Tot(M) one gets H^0(Σ̄, O(hC0) ⊗ J_{p,q,K}(w,y)) = 0. A standard deformation to the normal cone then transfers this local vanishing to the plane. Concretely, if the weighted local ideal has no nonzero global sections of the indicated degree on the normal surface, then there exist regular local parameters (u,v) at a point of P2 for which H^0(P2, O(3h) ⊗ J_{p,q,K}(u,v)) = 0. (F8, F9)
How this gives Nagata’s inequality Take p = 1, q = r and K = rm. Under the hypothesis K ≥ 3 h √r (equivalently rm ≥ 3 h √r), the previous steps yield local parameters on P2 for which the weighted ideal (u^r, v)^m has no nonzero degree-3h section. Lemma 2.2 provides a flat deformation turning this weighted monomial scheme into r distinct multiplicity-m fat points while preserving injectivity of the restriction map. Thus no degree-3h form can vanish to order at least m at r general points when 9 h^2 ≤ r m^2. A cube trick then extends the statement to arbitrary positive integers d,m, giving the inequality d > m √r in the asserted generality. (F10, F5)
Related weighted and valuative results A weighted (p,q) version of the construction yields a valuative statement: for coprime positive integers p,q with q > 9 p and for very general local parameters at a fixed point, every nonzero section F of O(d) satisfies ν(F) < d √(pq) for all d ≥ 1, while the asymptotic invariant μ̂(ν) equals √(pq). In particular, the supremum defining μ̂(ν) is not attained by any finite-degree form, even when pq is a square. (Here ν denotes the valuation measuring weighted vanishing; μ̂ is its asymptotic supremum.) (F11, F12, F18)
A geometric corollary For the very general parameters of the valuative statement, the same calculations determine the nef and effective cones of the weighted blow-up Y of P2 at one point. Concretely, the nef cone is R_{≥0}H + R_{≥0}D and the effective cone is R_{≥0}E + R_{≥0}D, where D = H − √(pq)E is strictly nef with D^2 = 0. No nonzero effective R-divisor lies on the ray of D, and the Cox ring of Y is not finitely generated. The theta-function calculation underlies this description when q/p > 9. (F14, F15)
Limitations and caveats The statements require “very general” choices. For Theorem A this means configurations outside a countable union of proper algebraic subsets. For the valuative/weighted results the parameters are chosen very generally in a fixed finite-dimensional jet space. These generality conditions are part of the hypotheses. (F16)
The restriction r ≥ 10 is necessary. Nine points always lie on a nonzero cubic, so the inequality cannot hold for r ≤ 9. (F17)
Although μ̂(ν) equals √(pq) asymptotically, no nonzero finite-degree form reaches the ratio ν(F)/deg(F) = √(pq) for the very general parameters; the supremum is not attained. (F18)
Summary The argument uses a sequence of reductions: deformation to the normal cone, collision into a monomial scheme, theta-function asymptotics as a cubic degenerates, interpolation duality, and a final combinatorial interpolation check. Together these reductions imply Nagata’s inequality d > m √r in the stated range. The key inputs are Proposition 2.4 (combinatorial interpolation), Theorem B (vanishing on the normal ruled surface), and Proposition 4.1 (transfer to P2). The numerical and generality hypotheses—r ≥ 10 and very general point configurations (and, in weighted statements, q/p > 9 and very general local parameters)—are essential to the conclusions. (F4, F8, F9, F16, F17)