at P we see that g ∗ f = f . It follows that g ∗ is trivial on all of k(X ) and hence
g is trivial on all of X . From the format of the matrix for g ∗ it follows that the 2 ∗ action on GLk MP (X )/MP (X ) ⋍ k is trivial on elements of order p, so that the map GP → k ∗ has cyclic image and the kernel is a p-group. It then follows that GP ⋍ Cm ⋉ H where H is a p-group and that eπ (P ) = m × q where q is some p-power. For g ∈ GP we call the image of g ∗ in k ∗ the rotation number of g at P.
2.6. Cyclic n-gonal curves. A cyclic n-gonal curve has a cyclic G = Cn
action for which the genus of X /G is zero. A defining equation is of the form y n = f (x), where f (x) is a rational function. The curve can be put in a canonical form: y n = (x − a1 )n1 · · · (x − at )nt , where n, n1 , . . . , nt ∈ Z+ , a1 , . . . , at ∈ k satisfy: (1) a1 , . . . , at are distinct, (2) 0 < ni < n, (3) n divides n1 + · · · + nt (4) gcd(n1 , . . . , nt ) = 1. The conditions 1 and 2 simplify the model and eliminate degeneracies, condition 3 ensures that curve is not ramified over ∞, and condition 4 ensures that curve is irreducible. The curve needs to be normalized to make it smooth. There is a G-action, where G = Un = {u ∈ k : un = 1}, given by (x, y) → (x, uy). We must assume the p does not divide n otherwise the field extension will be inseparable. The quotient map is (x, y) → x, and is branched over a1 , . . . , at . The local equation is y n = b(x)(x − ai )ni , where b(ai ) 6= 0. There are di = gcd(n, ni ) branches of the curve meeting at (ai , 0), and in the normalization the ramification degree is mi = n/di . For a discussion of the p = 0 case, which extends in many ways to the p > 0 case see [18],[16]. In the complex case, the vector (n1 , . . . , nt ) is a generating vector (see Section 6.1) for G = Cn by the conditions 2, 3, 4. In the p > 0 case (n1 , . . . , nt ) is a classifier for the Cn action and may be thought of as a generating vector for Galois covers of Y = P1 (k), ramified exactly over B = {a1 , . . . , at }, with Galois group G = Cn . It plays the role of generating vector, though it is not constructed topologically. We will look at such curves in more detail in Section 5.
3. Weierstrass Gap Theorem and Weierstrass points
We assume that the reader is familiar with basic definitions on divisors on curves. For a short introduction see [27] in this volume or [70]. Let P be a point on Xg and consider the vector spaces L(nP ) for n = 0, 1, . . . , 2g− 1. These vector spaces contain functions with poles only at P up to a specific order. This leads to a chain of inclusions L(0) ⊆ L(P ) ⊆ L(2P ) ⊆ · · · ⊆ L((2g − 1)P ) with a corresponding non-decreasing sequence of dimensions ℓ(0) ≤ ℓ(P ) ≤ ℓ(2P ) ≤ · · · ≤ ℓ((2g − 1)P ). The following proposition shows that the dimension goes up by at most 1 in each step.