Text size

Section 30

Exclusion of all centres and completion of the proof

Lemma 30.1 (Curve centres in the exceptional divisor). Suppose that ℓ\ell is of the second kind. Every prime divisor over XX whose centre is a curve contained in EE has positive beta invariant.

Proof. Let CC be its centre. A ruling of E→ℓE\to\ell meets the negative section ss, which lies in f−1(y0)f^{-1}(y_0). If this fibre is reduced, Lemma 25.1 excludes such a centre; if it is nonreduced, Corollary 28.6 does so.

Otherwise C∼as+bgC\sim as+bg with integers a≥1a\ge1 and b≥0b\ge0, so C−sC-s is linearly equivalent to an effective divisor. Monotonicity of the surface volume and Lemma 26.2 give S(WE;C)≤S(WE;s)=7172<1.S(W^E;C)\le S(W^E;s)=\frac{71}{72}<1. At the generic point of CC, the only divisorial valuation of the surface EE is a positive multiple of ordC\operatorname{ord}_C. Thus its refined threshold there is at least 72/7172/71. Abban–Zhuang adjunction along the smooth divisor EE, with SX(E)=5/9S_X(E)=5/9, gives AX(Q)SX(Q)≥min{95,7271}>1\frac{A_X(Q)}{S_X(Q)}\ge \min\left\{\frac95,\frac{72}{71}\right\}>1 for every prime divisor QQ with centre CC. This proves the assertion. ◻

Lemma 30.2 (Transverse intersection with a smooth conic). Suppose that ℓ\ell is of the second kind and that p∈E\sp\in E\setminus s lies on a smooth conic fibre. In the notation of Lemma 29.1, assume that (cF)p=1(cF)_p=1. Every prime divisor QQ over XX with centre pp satisfies βX(Q)≥3172ordQ(S)>0.\beta_X(Q)\ge\frac{31}{72}\operatorname{ord}_Q(S)>0.

Proof. For the refinement of the surface series by FF, Lemma 24.2 gives 𝒮F=29/48\mathcal S_F=29/48 and 𝒯p=1\mathcal T_p=1. Applying surface adjunction first shows that the threshold of WSW^S at pp is at least one. Now SX(S)=41/72<1S_X(S)=41/72<1, so the strict adjunction inequality of Lemma 21.2 gives the displayed lower bound. Since the centre is p∈Sp\in S, the order of SS is positive. ◻

Proof of Theorem 23.1. Suppose first that ℓ\ell is of the first kind, and let QQ be a prime divisor over XX. If its centre is contained in EE, Lemma 26.1 gives βX(Q)>0\beta_X(Q)>0. Otherwise choose a point of its centre outside EE. According as its conic fibre is smooth, reduced and singular, or nonreduced, Lemma 24.2, Lemma 24.3, or Corollary 28.6 gives a local threshold strictly greater than one. Thus βX(Q)>0\beta_X(Q)>0 in every case, proving K-stability for first-kind lines.

Now suppose that ℓ\ell is of the second kind. Let QQ be any prime divisor over XX. If its centre is a divisor on XX, Lemma 24.1 gives βX(Q)>0\beta_X(Q)>0.

Suppose next that its centre is a curve. If the curve is contained in EE, apply Lemma 30.1. Otherwise take a point of the curve outside EE. For a smooth conic fibre, Lemmas 23.4 and 24.2 give local threshold at least 9/89/8. For a reduced singular fibre away from y0y_0, Lemma 24.3 gives at least 36/3536/35; for the reduced fibre over y0y_0, use Lemma 25.1. For a nonreduced fibre, Corollary 28.6 gives at least 144/127144/127. In every case this threshold is strictly greater than one and therefore excludes βX(Q)≤0\beta_X(Q)\le0.

It remains to consider a point centre pp. If p∉Ep\notin E, the same four estimates apply. If p∈Ep\in E lies on a singular conic fibre, the reduced-fibre lemmas and the double-fibre corollary again apply. The only case left is a point of EE on a smooth conic. Such a point is outside ss, because the fibre containing ss is singular. Choose SS as in Lemma 29.1. If (cF)p=1(cF)_p=1, Lemma 30.2 gives strict positivity. If (cF)p=2(cF)_p=2, the coordinate normalization of Lemma 29.1 permits application of Lemma 29.7, which gives a local threshold strictly greater than one. Thus βX(Q)>0\beta_X(Q)>0 for every prime divisor QQ over XX. The valuative criterion proves K-stability. ◻