Text size

Section 90

Two divisor refinements

We first compute two expected vanishing orders on the original threefold. These calculations also fix the normalizations used later.

Lemma 90.1. The divisor B=L−T=T+3H−EB=L-T=T+3H-E is globally generated. The positive and fixed parts of L−uDL-uD are uPD(u)ND(u)0≤u≤1L−uD01≤u≤3/22T+(3−2u)H(u−1)E.\begin{array}{c|c|c} u&P_D(u)&N_D(u)\\ \hline 0\le u\le1&L-uD&0\\ 1\le u\le3/2&2T+(3-2u)H&(u-1)E. \end{array} For L−uTL-uT they are uPT(u)NT(u)0≤u≤1L−uT01≤u≤2(2−u)B+(u−1)H(u−1)D.\begin{array}{c|c|c} u&P_T(u)&N_T(u)\\ \hline 0\le u\le1&L-uT&0\\ 1\le u\le2&(2-u)B+(u-1)H&(u-1)D. \end{array} The displayed positive parts are nef; the respective pseudoeffective thresholds are 3/23/2 and 22.

Proof. The ideal of CC is generated, after twisting by 𝒪M(1,3)\mathcal O_M(1,3), by the equation of SS times sections of 𝒪M(1,1)\mathcal O_M(1,1) and by two lifts of x(uf+vg)x(uf+vg) and y(uf+vg)y(uf+vg) from SS. This proves the assertion about BB. All the proposed positive parts are nonnegative linear combinations of LL, BB, TT and HH. In the first table, the first interval interpolates between LL and 2T+H2T+H. For u>1u>1, restriction to a fibre of E→CE\to C forces the component (u−1)E(u-1)E. After its removal, the remaining divisor is the pullback written in the table, and has no fixed divisorial component.

In the second table, L−uTL-uT restricts to (1−u)ℓ1+ℓ2(1-u)\ell_1+\ell_2 on DD. For u>1u>1, its negative degree on the appropriate ruling forces multiplicity at least u−1u-1 along DD, because D|D=−ℓ1−ℓ2D|_D=-\ell_1-\ell_2. Removing this multiplicity gives (2−u)B+(u−1)H(2-u)B+(u-1)H, which is globally generated when its coefficients are integral. The same conclusions for rational and real coefficients follow by taking multiples and continuity. Finally, beyond the stated thresholds the pushforward to MM has a negative coefficient of HH, respectively TT, and cannot be effective. ◻

Lemma 90.2 (Separate computations of SX(D)S_X(D) and SX(T)S_X(T)). One has SX(D)=3140,SX(T)=6980.S_X(D)=\frac{31}{40},\qquad S_X(T)=\frac{69}{80}.

Proof. For the first divisor, the intersection identities give PD(u)3={20−6u−6u2−2u3,0≤u≤1,6(3−2u)2,1≤u≤3/2.P_D(u)^3= \begin{cases} 20-6u-6u^2-2u^3,&0\le u\le1,\\ 6(3-2u)^2,&1\le u\le3/2. \end{cases} Consequently SX(D)=120(∫01(20−6u−6u2−2u3)du+∫13/26(3−2u)2du)=120(292+1)=3140.S_X(D)=\frac1{20} \left(\int_0^1(20-6u-6u^2-2u^3)\,du +\int_1^{3/2}6(3-2u)^2\,du\right) =\frac1{20}\left(\frac{29}{2}+1\right)=\frac{31}{40}. For the fibre divisor, the two volume polynomials are PT(u)3={20−12u,0≤u≤1,(u−2)(u2+2u−11),1≤u≤2.P_T(u)^3= \begin{cases} 20-12u,&0\le u\le1,\\ (u-2)(u^2+2u-11),&1\le u\le2. \end{cases} Thus SX(T)=120(∫01(20−12u)du+∫12(u−2)(u2+2u−11)du)=120(14+134)=6980.S_X(T)=\frac1{20} \left(\int_0^1(20-12u)\,du +\int_1^2(u-2)(u^2+2u-11)\,du\right) =\frac1{20}\left(14+\frac{13}{4}\right)=\frac{69}{80}. Both computations use volume on XX, not on a weighted extraction. ◻

Lemma 90.3 (Points on the distinguished surface). For every p∈Dp\in D, one has δp(X)≥80/73\delta_p(X)\ge80/73.

Proof. Take the ruling Z∈|ℓ2|Z\in|\ell_2| through pp. On DD, the moving parts in Lemma 90.1 have classes (1+u)(ℓ1+ℓ2)(0≤u≤1),2ℓ1+(6−4u)ℓ2(1≤u≤3/2).(1+u)(\ell_1+\ell_2)\quad(0\le u\le1),\qquad 2\ell_1+(6-4u)\ell_2\quad(1\le u\le3/2). The fixed restriction in the second interval is (u−1)C(u-1)C. The volume of aℓ1+bℓ2a\ell_1+b\ell_2 is 2ab2ab when a,b≥0a,b\ge0. The first refinement therefore gives S(WD;Z)=320∫01∫01+u2(1+u)(1+u−v)dvdu+320∫13/2∫06−4u4(6−4u−v)dvdu=6180.\begin{aligned} S(W^D;Z) &=\frac3{20}\int_0^1\int_0^{1+u} 2(1+u)(1+u-v)\,dv\,du\\ &\quad+\frac3{20}\int_1^{3/2}\int_0^{6-4u} 4(6-4u-v)\,dv\,du =\frac{61}{80}. \end{aligned} For the flag D⊃Z⊃{p}D\supset Z\supset\{p\} the moving contribution is 320(∫01∫01+u(1+u)2dvdu+∫13/2∫06−4u4dvdu)=6980.\frac3{20}\left( \int_0^1\int_0^{1+u}(1+u)^2\,dv\,du +\int_1^{3/2}\int_0^{6-4u}4\,dv\,du\right)=\frac{69}{80}. Since C→QC\to Q is an isomorphism, CC meets ZZ transversely once. The fixed contribution is zero for p∉Cp\notin C, and otherwise equals 620∫13/2∫06−4u2(u−1)dvdu=120.\frac6{20}\int_1^{3/2}\int_0^{6-4u}2(u-1)\,dv\,du =\frac1{20}. All discrepancies in this smooth surface flag are one. The flag inequality, together with Lemma 90.2, now gives δp(X)≥min{4031,8061,8073}=8073.\delta_p(X)\ge \min\left\{\frac{40}{31},\frac{80}{61},\frac{80}{73}\right\} =\frac{80}{73}. This reproduces [3, Lemma 5.68] with the two restriction integrals and the fixed contribution written separately. ◻

Lemma 90.4 (Smooth fibres). If pp lies on a smooth fibre TT, then δp(X)>1\delta_p(X)>1. More precisely, the bound is 80/7380/73 on DD, and at least 10/910/9 away from DD.

Proof. Only p∉Dp\notin D remains. Set CT=D|TC_T=D|_T. The moving restrictions are −KT-K_T for 0≤u≤10\le u\le1 and −KT−(u−1)CT-K_T-(u-1)C_T for 1≤u≤21\le u\le2; the fixed restrictions are supported on CTC_T. For any divisor FF over TT whose centre contains pp, the fixed contribution is zero. Monotonicity of volume gives S(WT;F)≤310∫0∞vol(−KT−vF)dv=65ST(F).S(W^T;F)\le\frac3{10}\int_0^\infty \operatorname{vol}(-K_T-vF)\,dv =\frac65 S_T(F). A smooth quartic del Pezzo surface satisfies δ(T)≥4/3\delta(T)\ge4/3 [3, Lemma 2.12]. Hence S(WT;F)≤(9/10)AT(F)S(W^T;F)\le(9/10)A_T(F). The first refinement has expected order 69/8069/80, so the Abban–Zhuang inequality yields δp(X)≥min{80/69,10/9}=10/9.\delta_p(X)\ge\min\{80/69,10/9\}=10/9. ◻