Text size

Section 107

Completion of the proof for Family 3.6

Proof of Theorem 103.1. Suppose that βX(F)≤0\beta_X(F)\le0 for a prime divisor FF over XX. By Proposition 104.6, its centre is contained in a double conic 2ℓ2\ell. Take the weighted extraction σ:Y→X\sigma:Y\to X constructed above. It is plt, its different is D/2D/2, and its actual refinement has threshold at least one by Proposition 106.10. Abban–Zhuang adjunction in its strict form gives βX(F)≥ordF(G)(AX(G)−SX(G))=2188ordF(G)>0.\beta_X(F)\ge\operatorname{ord}_F(G) \left(A_X(G)-S_X(G)\right) =\frac{21}{88}\operatorname{ord}_F(G)>0. The last inequality holds because cX(F)⊂ℓc_X(F)\subset\ell, so the centre on YY is contained in the exceptional divisor. This is the required contradiction.

For clarity, the strict inequality retains more information than the minimum bound for local delta invariants. Choose a basis compatible with the FF- and GG-filtrations and restrict its moving part to GG. The refined bound and inversion of adjunction give, after taking limits, SX(F)≤AX(F)−AX(G)ordF(G)+SX(G)ordF(G).S_X(F)\le A_X(F)-A_X(G)\operatorname{ord}_F(G) +S_X(G)\operatorname{ord}_F(G). This is precisely the displayed positive remainder. Thus a refined threshold equal to one causes no gap in the argument. ◻

Remark 107.1 (The role of the section calculation). No Zariski decomposition on the weighted threefold YY has been computed or assumed. The prime initial Cox equation identifies the valuation filtration. Its character gives SX(G)=243/88S_X(G)=243/88, and the exact binary-space count identifies every refined section, including cancellations. The remaining Zariski decompositions are on the blow-up of ℙ1×ℙ1\mathbb P^1\times\mathbb P^1 at four points. When marked points coalesce, a separate degeneration keeps the different smooth and gives the threshold one. In particular, the nonnormal sextic fibre occurring when Λ\Lambda passes through a quadric-cone vertex is included in the proof; it is never treated as a Du Val surface.