Show that higher‑genus dominant saddles are exactly branched covers of the sphere

Establish that, for any genus g≥2, the dominant saddle geometries in the high‑energy fixed‑angle limit of the closed‑string four‑point amplitude are precisely the (g+1)‑fold branched covers of the sphere branched at the four vertex insertions, and prove that no other saddle geometries with equal or smaller real part of the action exist.

Background

Motivated by the one‑loop analysis, the authors hypothesize that at genus g the leading saddles are obtained by lifting the tree‑level solution to a (g+1)‑fold branched cover. This would imply a universal rescaling of the on‑shell action and predictable phase structure.

However, a general proof excluding other candidate saddles at higher genus is not known, and establishing such a classification would generalize the one‑loop picture.

References

In particular, it is natural to assume that the dominant saddles are given by surfaces $\mathcal{M}_{g,4}$ which are $g+1$-fold covers of the tree-level saddle, branched over four points where the vertex operators are inserted (even though we are not aware of a proof that there are no other candidate saddles).

Precision asymptotics of string amplitudes  (2601.09707 - Baccianti et al., 14 Jan 2026) in Section 6, Conclusion (Higher genus paragraph)