Cor: (Hurwitz bound). Suppose
X has genus
≥2 and finitely many automorphisms. Then
∣Aut(X)∣≤84(g−1).
Proof. let fi be the number of points of ramification index ri over a point pi in the base. Orbit stabilizer theorem says that ∣Aut(X)∣=firi. Let h denote the genus of X/Aut(X). Riemann-Hurwitz says that
2g(X)−2=∣Aut(X)∣⋅(2h−2)+i,j∑(ei,j−1)=∣Aut(X)∣⋅(2h−2)+i=1∑bfi(ri−1)=∣Aut(X)∣⋅(2h−2−i=1∑b(1−ri1)). Set
c=(2h−2−∑i(1−ri1)). We'll come up with a lower bound on
c by examining a few cases:
Case 1: h≥2. Then c≥2 and ∣Aut(X)∣≤g−1.
Case 2: h=1. Then b≥1 (remember, b is the number of branch points) and c≥21. Thus ∣Aut(X)∣≤4(g−1).
Case 3: h=0.
Case 3.1: If b≥5, then c≥21 and so ∣Aut(X)≤4(g−1).
Case 3.2: If b=4 then not all ri can be equal to 2. Then
c≥−223=6g+4−12=61⟹∣Aut(X)∣≤12(g−1).
Case 3.3: The case that b≤2 is not possible since g is at least 2.
Case 3.4: If b=3 then suppose r1≤r2≤r3 without loss of generality. Now MORE subcases:
Case 3.4.1: If r1≥3, then not all ri can be equal to 3, so
c≥−2+32+32+43=12−24+16+9=121, from which it follows that
∣Aut(X)∣≤24(g−1).
Case 3.4.2: If r1=2 then r2>4 and r3≥5. Then
c≥−2+21+43+54=20−40+10+15+16=201 so
∣Aut(X)∣≤40(g−1).
Case 3.4.3: If r1=2, r2=3 and then r3≥7. Then
c≤−2+21+43+76=421 so
∣Aut(X)∣≤84(g−1), which is exactly the Hurwitz bound.
So in all cases, if X is a genus ≥2 curve, then ∣Aut(X)∣≤84(g−1).
Lemma: If
X has genus
g≥1 and
L has degree
2g−2 then
h0(L)≤g and equality holds if and only if
L≅KX.
Proposition: If
X is a hyperelliptic curve with
g≥2 then
KX≅L⊗g−1 where
L is a
g21.
Proof. L induces a map
φL:X→P1. Consider the Veronese embedding of
P1, i.e. the map
P1→Pg−1 induced by
OP1(g−1):
[x,y]↦[x3,x2y,xy2,y3]. Then
φL∗OP1(1)≅L and
OP1(g−1)≅OP1(1)⊗g−1. Thus
φ∘φL:X→Pg−1 given by
L⊗g−1. In particular,
h0(L⊗g−1) and
deg(L⊗g−1)=2g−2, hence
Lg−1≅KX.
Theorem: Let g≥2. Then KX is base point free and
very ample if X is not hyperelliptic
a 2:1 cover of a rational normal curve in Pg−1 if X is hyperelliptic.
Proof. Let p∈X.
h0(KX−p)−h0(OX(p))=2g−3−g+1=g−2. We previously saw that
h0(OX(p))=1 in this case (I forgot why this is true) so
h0(KX−p)=g−1=h0(KX)−1.
Since h0(OXO(p+q)=1
h0(Kx−p−q)−h0(OXO(p+q))=2g−4−g+1=g−2⟹h0(Kx−p−q)=g−2.
This was the previous proposition.
Proposition: If
X is a hyperelliptic curve of genus
≥2 then the
g21 is unique.
Proof. Consider
φKX:X→Pg−1. By the previous proposition, this is a
2:1 cover of its image which is a rational curve. Conversely, give any
g21 on
X and it defines this cover. This implies the
g21 is unique.