Reflection principle for strange-duality sections

Determine whether, whenever a theta line bundle \(\Theta\) has negative intersection \((\Theta,\Delta)=-k<0\) with an effective divisor \(\Delta\) of square \(-2\), every theta section produced by the strange duality map lies in \(s_\Delta^k H^0(M,s_\Delta(\Theta))\), and whether the rank of the strange duality map equals the dimension of the section space obtained by successive reflections in such divisors.

Background

The three counterexamples share a mechanism involving an effective square-−2-2 divisor on a moduli space and a theta line bundle with negative degree on that divisor. The strange-duality sections vanish along the divisor to an order greater than the fixed multiplicity forced by the complete linear system.

Reflecting the theta class in the square-−2-2 divisor produces a better-behaved line bundle in the examples. The authors ask whether this reflection procedure universally controls both the image of the strange duality map and its rank.

References

Suppose that a theta line bundle \Theta fails to be nef because of an effective divisor \Delta of square -2 with (\Theta,\Delta)=-k<0. Do all theta sections produced by the strange duality map then lie in s_\Deltak\cdot H0(M,s_\Delta(\Theta))? More generally, is the rank of SD equal to the dimension of the space of sections of the class obtained from \Theta by successive reflections in such divisors?

— A refinement of the strange duality conjecture on K3 surfaces  (2610.21272 - Zhang et al., 15 Oct 2026) in Question 5.3, Section 5.2