How Many Rational Curves Hide on K3 Surfaces?

K3 surfaces are among the most captivating objects in algebraic geometry: smooth, complex, with zero curvature everywhere. A long‑standing question is how many rational curves—those that can be parametrized by a line—can sit on a K3 of a given "size".

A recent paper (arXiv:2512.06145) solved this for degree‑d curves with d>1 on a K3 of degree 2n, provided n is much larger than d. The authors proved that the maximal number of rational curves is governed by Miyaoka's bound. When n is odd and sufficiently large, the bound is sharp: the curves arrange themselves in very regular configurations, like a geometric dance.

Picture a K3 as an endless dance floor; rational curves are the dancers. Miyaoka's bound tells how many dancers can fit without colliding. This work shows that, if the floor is big enough, the dancers can fill every available spot, but only in specific patterns.

The result is more than a number; it maps how the curves intertwine, opening new avenues to study symmetry and topology of K3 surfaces. In the future, these insights could also impact theoretical physics, where K3s appear in string theory models.

Miyaoka's bound, introduced in the 1970s, provides an inequality between the number of rational curves and the topological invariants of a surface of general type. Technically, for a surface X of general type with c_1^2(X) and c_2(X) denoting the square of the first Chern class and the second Chern class respectively, Miyaoka proves that 3c_2(X) ≥ c_1^2(X). When X is a K3 surface, c_1^2=0 and c_2=24, so the inequality is saturated. However, the presence of rational curves introduces extra terms in the Noether‑Bogomolov‑Miyaoka‑Yau formula, since each curve C of genus g and self‑intersection C^2 contributes (2g‑2)+C^2. For rational curves (g=0) of degree d on a K3 of degree 2n, the self‑intersection is C^2 = 2d^2 n ‑ 2d, derived from the adjunction formula. The paper shows that if n≫d, the sum of these contributions reaches exactly the limit imposed by Miyaoka, turning the inequality into an equality. This occurs only when n is odd, because parity affects the ability to realize curve configurations that satisfy the integrality conditions of the Picard lattice of the K3. Moreover, the optimal configurations correspond to sets of curves forming ADE-type lattices, linked to the symmetries of simple Lie groups. Such lattices are also crucial in physics, where K3 surfaces with ADE configurations appear in constructing gauge‑theory models in string theory. The work therefore opens new avenues to connect classical algebraic geometry with deeper symmetry structures, suggesting possible generalizations to surfaces with different Chern classes or to higher‑dimensional varieties.