Algebraic Curves Lecture 22

The 22nd lecture of algebraic curves by Karl Christ
  1. Extremal Curves
  2. Interlude: Rational normal curves and rational normal scrolls
  3. Rational normal scrolls

I missed the previous lecture due to an unfortunate phone call.

Extremal Curves

d<2rd < 2r means you've got a non-special line bundle. Not sure where this showed up last time as I wasn't there.

Interlude: Rational normal curves and rational normal scrolls

Suppose C⊂PrC\subset \mathbb P^r is embedded by OP1(r)\mathcal O_{\mathbb P^1}(r),

[s:t]↦[sr:sr−1t:...:tr].\begin{aligned} [s:t] \mapsto [s^r:s^{r-1}t:...:t^r]. \end{aligned}

Remember:

Set

M=(x0x1x2…xr−1x1x2x3…xr).\begin{aligned} M = \begin{pmatrix} x_0 & x_1 & x_2 & \dots & x_{r-1} \\ x_1 & x_2 & x_3 & \dots & x_r \end{pmatrix}. \end{aligned}

The ideal of CC is defined by 2×22\times 2 minors of MM. There are (r2)\binom{r}{2} linearly independent quadrics cutting out CC.

φ:H0(Pr,OPr(2))→H0(C,OC(2)).\begin{aligned} \varphi:H^0(\mathbb P^r, \mathcal O_{\mathbb P^r}(2)) \to H^0(C, \mathcal O_C(2)). \end{aligned}

and

dim⁡(img⁡φ)=(r+22)−(r2)=(r+2)(r+1)−r(r−1)2=r2+3r+2−r2+r2=2r+1=h0(C,OC(2))\begin{aligned} \operatorname{dim}(\operatorname{img}\varphi) &= \binom{r+2}{2} - \binom{r}{2} \\ &= \frac{(r+2)(r+1) - r(r-1)}{2} \\ &= \frac{r^2 + 3r + 2 - r^2 + r}{2} \\ &= 2r + 1 = h^0(C,\mathcal O_C(2)) \end{aligned}

which implies that φ\varphi is surjective which happens if and only if CC is 2-normal.

Lemma: If C⊂PrC\subset \mathbb P^r is embedded by a non-special line bundle LL (i.e. C=φL(X)C = \varphi_L(X), h0(KX−L)=0h^0(K_X - L) = 0 then CC is projectively normal if and only if CC is 22-normal.

Proof.   Assume CC is 00, 11, 22-normal. We need to show that CC is ℓ\ell normal for any ℓ>3\ell > 3. Consider the exact sequence:

0→OC(1)→⋅HOC(2)→OC∩H(2)→0.\begin{aligned} 0 \to \mathcal O_C(1) \xrightarrow{\cdot H} \mathcal O_C(2) \to \mathcal O_{C\cap H}(2)\to 0. \end{aligned}
As is standard, take the long exact sequence on cohomology
H0(Pr,OPr(2))→H0(C,OC(2))→H0(C∩H,OC∩H(2))→H1(C,OC(1)).\begin{aligned} H^{0}(\mathbb P^r, \mathcal O_{\mathbb P^r}(2)) \to H^0(C, \mathcal O_C(2)) \to H^0(C\cap H, \mathcal O_{C\cap H}(2)) \to H^1(C, \mathcal O_C(1)). \end{aligned}
This means the points in C∩HC\cap H impose independent conditions on quadrics in Pr\mathbb P^r which in turn implies the points also impose independent conditions on all hypersurfaces of degree ℓ≥2\ell \geq 2.

Now consider the exact sequence of the ideal sheaf IC∩H\mathcal I_{C\cap H} given by

0→IC∩H(ℓ)→OPr(ℓ)→OC∩H(ℓ)→0.\begin{aligned} 0 \to \mathcal I_{C\cap H}(\ell) \to \mathcal O_{\mathbb P^r}(\ell) \to \mathcal O_{C\cap H}(\ell) \to 0. \end{aligned}
Then
H1(C,IC∩H(ℓ))=0\begin{aligned} H^1(C, \mathcal I_{C\cap H}(\ell)) = 0 \end{aligned}
for all ℓ≥2\ell\geq 2, since OPr(ℓ)→OC∩H(ℓ)\mathcal O_{\mathbb P^r}(\ell) \to \mathcal O_{C\cap H}(\ell) is surjective on global sections whenever ℓ≥2\ell \geq 2. We'd like to show that H1(C,IC∩H(1)=1H^1(C, \mathcal I_{C\cap H}(1) = 1 too. We have that
0→IC∩H(ℓ−1)→IC∩H(ℓ)→IC∩HH(ℓ)→0\begin{aligned} 0 \to \mathcal I_{C\cap H}(\ell - 1) \to \mathcal I_{C\cap H}(\ell) \to \mathcal I_{C\cap H}^H(\ell) \to 0 \end{aligned}
where IC∩H\mathcal I_{C\cap H} is the ideal sheaf of C∩HC\cap H inside of Pr\mathcal P^r and IC∩HH(ℓ)\mathcal I_{C\cap H}^H(\ell) is the ideal sheaf of C∩HC\cap H inside of H≅Pr−1H\cong \mathbb P^{r-1}. This exact sequence comes from the fact that...I'm not sure. Will need to consult the notes.
□\square
 

Thus, rational normal curves are projectively normal.

Theorem: (Griffths, Harris page 528 - 531).

  1. Through any r+3r+3 points in Pr\mathbb P^{r} in linear general position, there is a unique rational normal curve.

  2. Fix integers dd and rr such that r≥2r\geq 2 and d≥2r+3d\geq 2r + 3. Let Γ⊆Pr\Gamma \subseteq \mathbb P^r be dd points in linear general position. Then Γ\Gamma lines on a rational normal curve if and only if Γ\Gamma imposes 2r+12r + 1 conditions on quadrics.

Rational normal scrolls

The rational normal scrolls will be surfaces which contain our extremal curves. Fix a1a_1 and a2a_2 such that a1≥a2a_1\geq a_2, a1≥0a_1\geq 0 a2≥0a_2\geq 0 and a1+a2>r−1a_1 + a_2 > r - 1. Fix subspaces L1L_1 and L2L_2 such that L1L_1 and L2L_2 span Pr\mathbb P^r and dim⁡(Li)=ai\operatorname{dim}(L_i) = a_i. Let Ci⊆Ci⊆LiC_i \subseteq C_i \subseteq L_i be a rational normal curve and fix an isomorphism φ:C1→C2\varphi:C_1\to C_2. Then

Xa1,a2=⋃p∈C1{line connecting P to φ(P)}.\begin{aligned} X_{a_1, a_2} = \bigcup_{p\in C_1} \{\text{line connecting } P \text{ to } \varphi(P)\}. \end{aligned}

Example:

An Observation: Up to linear automorphisms of Pr\mathbb P^r this construction does not depend on the choices made other than a1a_1 and a2a_2.

Lemma:

  1. Xa1,a2X_{a_1,a_2} is a non-degenerate surface

  2. The degree of Xa1,a2X_{a_1,a_2} is a1+a2=r−1a_1+a_2 = r - 1

  3. Xa1,a2X_{a_1,a_2} is non-singular if a1≠0a_1 \neq 0.

Proof.  

  1. First, span⁡(ci)=Li\operatorname{span}(c_i) = L_i and span⁡(L1∪L2)=Pr\operatorname{span}(L_1\cup L_2) = \mathbb P^r. Any linear subspace H⊃Xa1,a2H\supset X_{a_1, a_2} needs to contain the Ci  ⟹  Li⊂HC_i \implies L_i \subset H, hence span⁡(L1∪L2)⊂H  ⟹  H=Pr\operatorname{span}(L_1\cup L_2) \subset H \implies H = \mathbb P^r.

  2. Xa1,a2∩H=C2+∑i=1a1LiX_{a_1,a_2} \cap H = C_2 + \sum_{i=1}^{a_1}L_i. Thus the degree of Xa1,a2X_{a_1,a_2} is equal to the degree of Xa1,a2∩HX_{a_1,a_2}\cap H which is a1+a2a_1 + a_2.

  3. Xa1,a2∩HpX_{a_1, a_2} \cap H_p. For any pp, I can choose HpH_p such that Xa1,a2∩HpX_{a_1,a_2} \cap H_p is smooth at pp as long as a1≠0a_1\neq 0. Then

deg⁡(H∩Xa1,a2)=r−1\begin{aligned} \operatorname{deg}(H\cap X_{a_1,a_2}) = r-1 \end{aligned}
which implies Xa1,a2∩H⊂HX_{a_1,a_2} \cap H\subset H is a rational normal curve if smooth.
□\square
 

Theorem:

What about the Picard group of Xa1,a2X_{a_1,a_2}? This is generated freely by the class LL of a line in the ruling and by HH the class of a hyperplane section. Note that you can write H=C1+a2⋅L=C2+a1⋅LH = C_1 + a_2 \cdot L = C_2 + a_1 \cdot L, so you can choose the class of one of the curves instead of HH if you prefer. Some intersection pairings:

Remarks: Remember

M=(x0x1x2…xr−1x1x2x3…xr).\begin{aligned} M = \begin{pmatrix} x_0 & x_1 & x_2 & \dots & x_{r-1} \\ x_1 & x_2 & x_3 & \dots & x_r \end{pmatrix}. \end{aligned}

Set

M′=(x0x1x2…xa1−1y0y1…ya2−1x1x2x3…xa1y0y1…ya2).\begin{aligned} M' = \begin{pmatrix} x_0 & x_1 & x_2 & \dots & x_{a_1-1} & y_0 & y_1 & \dots & y_{a_2 - 1} \\ x_1 & x_2 & x_3 & \dots & x_{a_1} & y_0 & y_1 & \dots & y_{a_2} \end{pmatrix}. \end{aligned}
  1. Xa1,a2X_{a_1, a_2} will be cut out by 2×22\times 2 minors of M′M', (r−12)\binom{r-1}{2} quadrics.

  2. The surface Xa1,a2X_{a_1,a_2} is projectively normal:

0→IXa1,a2→⋅HIXa1,a2(ℓ)→IXa1,a2∩HH(ℓ)→0.0 \to I_{X_{a_1,a_2}} \xrightarrow{\cdot H} I_{X_{a_{1},a_2}}(\ell) \to I_{X_{a_1,a_2}\cap H}^H(\ell) \to 0.
  1. Xa1,a2≅P(OP1(a1)⊕OP1(a2))≅F∣a2−a1∣X_{a_1,a_2}\cong \mathbb P(\mathcal O_{\mathbb P^1}(a_1)\oplus \mathcal O_{\mathbb P^1}(a_2))\cong \mathbb F_{|a_2 - a_1|} where F∣a2−a1∣\mathbb F_{|a_2 - a_1|} is a Hirzebruch surface.

©Isaac Martin. Last modified: April 05, 2024.