Text size

Section 53

The conic bundle and the remaining fibres

Family 2.16: conics on intersections of two quadrics

Let V⊂ℙ5V\subset\mathbb P^5 be a smooth intersection of two quadrics, let C⊂VC\subset V be a smooth conic, and let π:X=BlCV→V\pi\colon X=\mathop{\mathrm{Bl}}_CV\to V. These are the smooth members of Family 2.16. Put H=π*𝒪V(1)H=\pi^*\mathcal O_V(1), denote the exceptional divisor by EE, and set B=H−EB=H-E. Projection from the plane Π=⟨C⟩\Pi=\langle C\rangle gives the conic bundle g:X→ℙ2g\colon X\to\mathbb P^2 defined by |B||B|. Thus L=−KX=H+B=2H−EL=-K_X=H+B=2H-E and L3=22L^3=22.

The estimates for reduced fibres are already available in [28, Propositions 5.2.3, 5.2.4 and 5.2.6]. The remaining fibres are double lines. We blow up their scheme-theoretic fibre ideals, obtaining an exceptional surface F≅ℙ1×ℙ1F\cong\mathbb P^1\times\mathbb P^1 with different D/2D/2, where DD is a smooth curve of bidegree (2,1)(2,1). The new calculation identifies its refined sections with powers of the ideal of one point on DD. This is simpler than the length-two condition in Family 2.13, but the different must still be retained.

Theorem 53.1. Every smooth member XX of Family 2.16 satisfies δ(X)≥176/171\delta(X)\ge176/171. More precisely, if pp lies on a nonreduced conic fibre, then δp(X)≥88/85\delta_p(X)\ge88/85.

Lemma 53.2 (Intersection numbers and three expected orders). The intersection numbers are H3=4H^3=4, H2E=0H^2E=0, HE2=−2HE^2=-2, and E3=−2E^3=-2. Moreover, SX(H)=411,SX(E)=2344,SX(B)=1322.S_X(H)=\frac4{11},\qquad S_X(E)=\frac{23}{44},\qquad S_X(B)=\frac{13}{22}. Every prime divisor on XX has positive beta invariant.

Proof. The blow-up formulas give the intersection numbers: in particular, degNC/V=(−KV)C+2g(C)−2=2\deg N_{C/V}=(-K_V)C+2g(C)-2=2. Both HH and BB are nef. If T≠ET\ne E is a prime divisor and T∼aH+bBT\sim aH+bB, intersection with a general conic fibre and a general fibre of E→CE\to C gives a,b≥0a,b\ge0. Consequently, the effective cone is generated by EE and BB, and (aH−bE)3=4a3−6ab2+2b3.(aH-bE)^3=4a^3-6ab^2+2b^3. The divisors L−uHL-uH and L−uEL-uE are nef for 0≤u≤10\le u\le1 and are not pseudoeffective for u>1u>1. Their volume integrals are SX(H)=122∫01(4(2−u)3−6(2−u)+2)du=411,SX(E)=122∫01(32−12(1+u)2+2(1+u)3)du=2344.\begin{aligned} S_X(H)&=\frac1{22}\int_0^1 \bigl(4(2-u)^3-6(2-u)+2\bigr)\,du=\frac4{11},\\ S_X(E)&=\frac1{22}\int_0^1 \bigl(32-12(1+u)^2+2(1+u)^3\bigr)\,du=\frac{23}{44}. \end{aligned} For L−uBL-uB, the pseudoeffective threshold is two. The divisor is nef on [0,1][0,1]. On [1,2][1,2] its fixed part is (u−1)E(u-1)E, and the moving part is (2−u)H(2-u)H. Indeed, intersection with the moving fibres of E→CE\to C forces that multiplicity, and the remaining divisor is semiample. Thus SX(B)=122(∫01(4(2−u)3−6(2−u)(1−u)2+2(1−u)3)du+∫124(2−u)3du)=1322.S_X(B)=\frac1{22}\left( \int_0^1\bigl(4(2-u)^3-6(2-u)(1-u)^2+2(1-u)^3\bigr)\,du +\int_1^2 4(2-u)^3\,du\right)=\frac{13}{22}. For T≠ET\ne E, the integers a,ba,b above are not both zero, so T−HT-H or T−BT-B is effective. Monotonicity of volumes gives SX(T)≤13/22<1=AX(T)S_X(T)\le13/22<1=A_X(T). The calculation for EE treats the other prime divisor. ◻

Proposition 53.3 (The known reduced-fibre estimates). If the conic fibre through pp is reduced, then δp(X)≥176/171\delta_p(X)\ge176/171.

Proof. We use the three cited propositions of [28]. For a smooth fibre, unless it is the negative section of E≅𝔽2E\cong\mathbb F_2, they give 176/161176/161 off EE and 176/169176/169 on EE. For a reduced singular fibre, Proposition 5.2.4 gives 176/171176/171, including points on EE and the intersection of its two components. In the exceptional smooth-fibre case, Proposition 5.2.6 gives 22/1922/19. Their minimum is 176/171176/171. These are local estimates for all divisors whose centres contain pp; no automorphism-group assumption occurs in these propositions. ◻

53.1 A section ring adapted to the projection

Choose equations Q1,Q2Q_1,Q_2 of VV so that Q1Q_1 contains Π\Pi and Q2|ΠQ_2|_\Pi is the equation of CC. On Π\Pi use three coordinates, and use t0,t1,t2t_0,t_1,t_2 transverse to Π\Pi. In the blow-up of Π\Pi in ℙ5\mathbb P^5, write ti=eyit_i=e y_i, with [y0:y1:y2]∈ℙ2[y_0:y_1:y_2]\in\mathbb P^2. The variable ee has divisor class E=H−BE=H-B.

Lemma 53.4 (The projective-bundle presentation). Let T=ℙℙ2(𝒪⊕3⊕𝒪(1))T=\mathbb P_{\mathbb P^2}(\mathcal O^{\oplus3}\oplus\mathcal O(1)), in the quotient convention, with tautological class HH and base class BB. Then T=BlΠℙ5T=\mathop{\mathrm{Bl}}_\Pi\mathbb P^5, and X⊂TX\subset T is the complete intersection F1=Q1/e=0,F2=Q2=0,degF1=H+B,degF2=2H.F_1=Q_1/e=0,\qquad F_2=Q_2=0, \qquad \deg F_1=H+B,\quad\deg F_2=2H. For a,b≥0a,b\ge0, its spaces H0(X,aH+bB)H^0(X,aH+bB) are the (a,b)(a,b) pieces of R=ℂ[s,x,z,e,y0,y1,y2](F1,F2),variables,x,zey0,y1,y2(H,B)-degree(1,0)(1,−1)(0,1).R=\frac{\mathbb C[s,x,z,e,y_0,y_1,y_2]}{(F_1,F_2)},\qquad \begin{array}{c|ccc} \text{variable}&s,x,z&e&y_0,y_1,y_2\\\hline (H,B)\text{-degree}&(1,0)&(1,-1)&(0,1). \end{array} In particular, Nm=h0(X,mL)=(2m+1)(11m2+11m+6)6.N_m=h^0(X,mL)=\frac{(2m+1)(11m^2+11m+6)}6.

Proof. The three coordinates on Π\Pi and the products eyie y_i give the six sections of HH, identifying TT with the graph of the projection. The restrictions of Q1,Q2Q_1,Q_2 give the displayed equations. Near each point of CC, one coefficient of a transverse variable in Q1Q_1 is a unit: otherwise that point would be singular on Q1Q_1 and hence on VV. Locally eliminate that variable. The incidence equations then become the usual graph of the blow-up of the remaining regular codimension-two ideal. Away from CC the graph is already VV. This proves the complete-intersection assertion, including along the exceptional divisor.

For a≥0a\ge0, the push-forward of 𝒪T(aH+bB)\mathcal O_T(aH+bB) is ⨁d=0aSyma−d(ℂ3)⊗𝒪ℙ2(b+d),\bigoplus_{d=0}^a \operatorname{Sym}^{a-d}(\mathbb C^3)\otimes \mathcal O_{\mathbb P^2}(b+d), and the higher direct images vanish. Thus its sections are exactly the polynomial piece with the stated degrees. Twists with HH-coefficient −1,−2,−3-1,-2,-3 are acyclic on the projective-space fibres. In the Koszul resolution of XX, the additional twists of aH+bBaH+bB have degrees (a−1,b−1)(a-1,b-1), (a−2,b)(a-2,b), and (a−3,b−1)(a-3,b-1). Whenever their HH-coefficient is nonnegative, their push-forwards have base twists at least −1-1, hence no higher cohomology. Otherwise they are among the three acyclic relative twists just listed. Taking sections therefore gives the asserted quotient. Its Hilbert function on the diagonal, or Riemann–Roch with L3=22L^3=22, gives NmN_m. ◻