Algebraic Curves Lecture 18

The 18th lecture of algebraic curves by Karl Christ
  1. Continuing the Petri Map

Continuing the Petri Map

The Petri map is

μ0:H0(X,L)H0(X,KXL)H0(X,KX)\begin{aligned} \mu_0:H^0(X,L)\otimes H^0(X,K_X - L) \to H^0(X,K_X) \end{aligned}

defined by s1s2s1s2s_1 \otimes s_2 \mapsto s_1s_2. We saw that when XX has genus 3, then μ0\mu_0 is injective if and only if L≇g21L\not\cong g^1_2. Let's expand a bit more on this example:

Proposition: (Base point free pencil trick). Let LL be a line bundle on XX and s1,s2H0(X,L)s_1, s_2 \in H^0(X,L) be linearly independent global sections. Denote by VV the span {s1,s2}\{s_1,s_2\} and let FF be another locally free sheaf. Then the kernel of the tensor product map
VH0(X,F)H0(X,LF)\begin{aligned} V\otimes H^0(X,F)\to H^0(X,L\otimes F) \end{aligned}
is H0(X,FL1(B))H^0(X,F\otimes L^{-1}(B)) where BB is the base locus of VV.

Proof.   Pick an element s1t2s2t1ker(φ)s_1\otimes t_2 - s_2\otimes t_1 \in \operatorname{ker}(\varphi). I can write any element of the kernel like this because s1s_1 and s2s_2 form a basis for VV. Let si=sris_i = s\cdot r_i with sH0(X,OX(B))s \in H^0(X,\mathcal O_X(B)) a global section of LL vanishing on VV with r1V(B)L(B)r_1 \in V(-B)\subseteq L(-B). This is confusing, so here's a disambiguation:

By assumption, s1t2s2t1=0s_1\cdot t_2 - s_2 \cdot t_1 = 0. This implies r1t2r2t1=0r_1\cdot t_2 - r_2 \cdot t_1 = 0 where r1r_1 and r2r_2 do not have common zeroes. This in turn means tit_i vanishes along DiD_i.

Write τi=tiri\tau_i = \frac{t_i}{r_i}. Since tiH0(X,F)t_i \in H^0(X,F) and 1/riH0(X,L1(B))1/r_i \in H^0(X,L^{-1}(B)), τi\tau_i is a rational section of FL1(B)F\otimes L^{-1}(B) (the inverse of L(B)L(-B) is L1(B)L^{-1}(B)). In fact, it is a regular section, because the zeros of rir_i and tit_i cancel each other out. This means τiH0(X,FL1(B))\tau_i \in H^0(X, F\otimes L^{-1}(B)).

I now have sections τ1\tau_1 and τ2\tau_2, and I want to show they are equal. This follows since ti=riτit_i = r_i\tau_i and hence

    r1r2τ2t2r2r1τ1t1=0    τ1=τ2=:τ.\begin{aligned} &\implies r_1\cdot \underbrace{r_2 \cdot \tau_2}_{t_2} - r_2\cdot \underbrace{r_1\cdot \tau_1}_{t_1} = 0 \\ &\implies \tau_1 = \tau_2 =: \tau. \end{aligned}
In summary,
s1t2s2t1=sτ(r1r2r2r1)\begin{aligned} s_1\otimes t_2 - s_2 \otimes t_1 = s\cdot \tau \cdot (r_1\otimes r_2 - r_2 \otimes r_1) \end{aligned}
where τH0(X,FL1(B))\tau \in H^0(X,F\otimes L^{-1}(B)).

Conversely, if we're given a τH0(X,FL1(B))\tau \in H^0(X,F\otimes L^{-1}(B)), then set ti=τrit_i = \tau\cdot r_i. Then

s1t2s2t1=sτ(r1r2r2r1),\begin{aligned} s_1\otimes t_2 - s_2 \otimes t_1 = s\cdot \tau(r_1\otimes r_2 - r_2 \otimes r_1), \end{aligned}
which is clearly in the kernel of φ\varphi.
\square
 

Using this trick we get the following corollary regarding the injectivity of the Petri map:

Cor: If (L,V)(L,V) is a gd1g^1_d with 2d<g22d < g - 2, then the Petri map μ0\mu_0 for (X,L)(X,L) is not injective.
Proof.   We may assume that VV is basepoint free by subtracting base points. The map
φ:VH0(X,KXL)H0(X,KX).\begin{aligned} \varphi:V\otimes H^0(X,K_X - L) \to H^0(X,K_X). \end{aligned}
The kernel of φ\varphi is H0(X,KX2L)H^0(X,K_X - 2L) by the base point free pencil trick. The degree of KX2LK_X - 2L is 2g22d>2g2g+2=g2g - 2 - 2d > 2g - 2 - g + 2 = g. By Riemann-Roch,
h0(L)h0(KXL)=dg+1    h0(L)dg+1,\begin{aligned} h^0(L) - h^0(K_X - L) = d - g + 1 \implies h^0(L) \geq d - g + 1, \end{aligned}
so as soon as d>gd > g, we know that there is a global section of LL. Hence KX2LK_X - 2L has a global section, so ker(φ)=H0(X,KX2L)0\operatorname{ker}(\varphi) = H^0(X,K_X - 2L) \neq 0.
\square
 
Remark: The dimension of the locus of curves admitting a gd1g^1_d is
    2g2=d(2h2)+b    b=2g2+2d.\begin{aligned} &\phantom{\implies}2g - 2 = d\cdot (2h - 2) + b \\ &\implies b = 2g - 2 + 2d. \end{aligned}
Then b32g5+2d3g3b - 3 - 2g - 5 + 2d \geq 3g - 3 and so 2dg+22d \geq g + 2.
©Isaac Martin. Last modified: March 18, 2024.