ConceptComplete

Ramification

Ramification is the phenomenon where a morphism of curves fails to be a local isomorphism. It measures how fibers of a map degenerate, generalizing the branching behavior of multi-valued functions in complex analysis. The concept lies at the heart of the Hurwitz formula, the classification of covers, and has deep analogues in algebraic number theory.


Morphisms of curves and degree

Definition4.12Finite morphism of curves

Let CC and DD be smooth projective curves over an algebraically closed field kk, and let f:Cβ†’Df: C \to D be a nonconstant morphism. Then ff is automatically finite and surjective. It induces an extension of function fields fβˆ—:k(D)β†ͺk(C)f^*: k(D) \hookrightarrow k(C).

The degree of ff is the degree of the field extension:

deg⁑f=[k(C):k(D)].\deg f = [k(C) : k(D)].

The morphism ff is separable if the extension k(C)/k(D)k(C)/k(D) is separable, and purely inseparable if every element of k(C)βˆ–k(D)k(C) \setminus k(D) is purely inseparable over k(D)k(D). In general, deg⁑f=deg⁑sfβ‹…deg⁑if\deg f = \deg_s f \cdot \deg_i f where deg⁑s\deg_s is the separable degree and deg⁑i\deg_i is the inseparable degree.

For any point Q∈DQ \in D, the fiber fβˆ’1(Q)f^{-1}(Q) satisfies βˆ‘P↦QeP=deg⁑f\sum_{P \mapsto Q} e_P = \deg f, where ePe_P is the ramification index at PP.

ExampleProjection from a plane curve

Let CβŠ‚P2C \subset \mathbb{P}^2 be a smooth plane curve of degree dd, and Ο€:Cβ†’P1\pi: C \to \mathbb{P}^1 the projection from a point Qβˆ‰CQ \notin C. A generic line through QQ meets CC in dd points, so deg⁑π=d\deg \pi = d.

Concretely, for C:y2=x3+1C: y^2 = x^3 + 1 (a genus-1 curve of degree 3 in its Weierstrass embedding), the projection Ο€(x,y)=x\pi(x,y) = x has deg⁑π=2\deg \pi = 2: the generic fiber consists of the two points (x0,Β±x03+1)(x_0, \pm\sqrt{x_0^3 + 1}).

ExampleThe power map on β„™^1

The map f:P1β†’P1f: \mathbb{P}^1 \to \mathbb{P}^1 given by f([x:y])=[xn:yn]f([x:y]) = [x^n : y^n] (or t↦tnt \mapsto t^n in affine coordinates) has degree nn. The function field extension is k(t)β†ͺk(s)k(t) \hookrightarrow k(s) where sn=ts^n = t, which has degree nn.

The generic fiber over t0β‰ 0,∞t_0 \neq 0, \infty consists of nn distinct points (the nn-th roots of t0t_0). Over t0=0t_0 = 0 and t0=∞t_0 = \infty, the fiber has a single point with multiplicity nn.


Ramification index

Definition4.13Ramification index and ramification point

Let f:Cβ†’Df: C \to D be a nonconstant morphism of smooth projective curves, and P∈CP \in C a closed point with Q=f(P)∈DQ = f(P) \in D. Let tt be a uniformizer at QQ (i.e., a generator of the maximal ideal mQβŠ‚OD,Q\mathfrak{m}_Q \subset \mathcal{O}_{D,Q}).

The ramification index of ff at PP is

eP=vP(fβˆ—t)e_P = v_P(f^* t)

where vPv_P is the valuation at PP. Equivalently, ePe_P is the largest integer ee such that fβˆ—t∈mPef^* t \in \mathfrak{m}_P^e.

  • If eP=1e_P = 1, the morphism is unramified at PP.
  • If ePβ‰₯2e_P \geq 2, the point PP is a ramification point of ff.
  • The image Q=f(P)Q = f(P) of a ramification point is called a branch point.
  • The branch locus of ff is the set of all branch points in DD.

The fundamental relation for each Q∈DQ \in D:

βˆ‘P∈fβˆ’1(Q)eP=deg⁑f.\sum_{P \in f^{-1}(Q)} e_P = \deg f.

ExampleSquare root map

For f:P1β†’P1f: \mathbb{P}^1 \to \mathbb{P}^1, t↦t2t \mapsto t^2 (degree 2):

  • Over Q=a2β‰ 0,∞Q = a^2 \neq 0, \infty: the fiber is {a,βˆ’a}\{a, -a\}, two distinct points with ea=eβˆ’a=1e_a = e_{-a} = 1. Unramified.
  • Over Q=0Q = 0: the fiber is {0}\{0\} with e0=2e_0 = 2. Ramified! The uniformizer at Q=0Q = 0 is tt, and fβˆ—t=s2f^*t = s^2 where ss is the coordinate on the source. So v0(s2)=2v_0(s^2) = 2.
  • Over Q=∞Q = \infty: similarly e∞=2e_\infty = 2.

The branch locus is {0,∞}βŠ‚P1\{0, \infty\} \subset \mathbb{P}^1. The sum checks: ea+eβˆ’a=1+1=2=deg⁑fe_a + e_{-a} = 1 + 1 = 2 = \deg f (generic fiber), and e0=2=deg⁑fe_0 = 2 = \deg f (ramified fiber).

ExampleRamification of the cubic map

For f:P1β†’P1f: \mathbb{P}^1 \to \mathbb{P}^1, t↦t3t \mapsto t^3 (degree 3):

  • Over Qβ‰ 0,∞Q \neq 0, \infty: three distinct preimages, each with e=1e = 1.
  • Over Q=0Q = 0: single preimage P=0P = 0 with e0=3e_0 = 3. Totally ramified.
  • Over Q=∞Q = \infty: single preimage P=∞P = \infty with e∞=3e_\infty = 3. Totally ramified.

Branch locus: {0,∞}\{0, \infty\}. The ramification divisor is R=2β‹…[0]+2β‹…[∞]R = 2 \cdot [0] + 2 \cdot [\infty] with deg⁑R=4\deg R = 4.

ExampleChebyshev polynomials

The degree-nn Chebyshev polynomial Tn:P1β†’P1T_n: \mathbb{P}^1 \to \mathbb{P}^1 satisfies Tn(cos⁑θ)=cos⁑(nΞΈ)T_n(\cos\theta) = \cos(n\theta). As a map of degree nn:

  • Over Q=Β±1Q = \pm 1: the preimages of 11 are cos⁑(2Ο€k/n)\cos(2\pi k/n) for k=0,…,nβˆ’1k = 0, \ldots, n-1, but some coincide. Specifically, TnT_n has ramification index 22 at each of cos⁑(Ο€k/n)\cos(\pi k/n) for k=1,…,nβˆ’1k = 1, \ldots, n-1 (the interior critical points).
  • Over generic QQ: nn distinct preimages.

For T2(t)=2t2βˆ’1T_2(t) = 2t^2 - 1: branch points at Β±1\pm 1. Over Q=1Q = 1: preimages at t=Β±1t = \pm 1 with e=1e = 1 each (since T2(Β±1)=1T_2(\pm 1) = 1 and T2β€²(Β±1)=Β±4β‰ 0T_2'(\pm 1) = \pm 4 \neq 0). Over Q=βˆ’1Q = -1: single preimage t=0t = 0 with e=2e = 2 (since T2(0)=βˆ’1T_2(0) = -1 and T2β€²(0)=0T_2'(0) = 0). So the branch locus is {βˆ’1}\{-1\}.


Ramification divisor and the different

Definition4.14Ramification divisor

Let f:C→Df: C \to D be a separable morphism of smooth projective curves. The ramification divisor of ff is

R=βˆ‘P∈C(ePβˆ’1)β‹…P.R = \sum_{P \in C} (e_P - 1) \cdot P.

This is an effective divisor supported on the ramification points of ff. Its degree is

deg⁑R=βˆ‘P∈C(ePβˆ’1).\deg R = \sum_{P \in C} (e_P - 1).

In the tame case (char⁑k∀eP\operatorname{char} k \nmid e_P for all PP), this coincides with the different divisor. In the wild case, the different may be strictly larger (see below).

Definition4.15Different and discriminant

More generally, for a finite separable morphism f:C→Df: C \to D, the different DC/D\mathfrak{D}_{C/D} is the effective divisor on CC defined by

DC/D=βˆ‘P∈CdPβ‹…P\mathfrak{D}_{C/D} = \sum_{P \in C} d_P \cdot P

where dP=vP(D)d_P = v_P(\mathfrak{D}) is computed from the local different of the extension OD,f(P)β†ͺOC,P\mathcal{O}_{D,f(P)} \hookrightarrow \mathcal{O}_{C,P} of discrete valuation rings.

  • Tame ramification (char⁑k∀eP\operatorname{char} k \nmid e_P): dP=ePβˆ’1d_P = e_P - 1.
  • Wild ramification (char⁑k∣eP\operatorname{char} k \mid e_P): dPβ‰₯ePd_P \geq e_P, strictly greater than ePβˆ’1e_P - 1.

The discriminant is the norm: Ξ”C/D=fβˆ—DC/D\Delta_{C/D} = f_* \mathfrak{D}_{C/D}, a divisor on DD.


Tame versus wild ramification

RemarkTame vs wild ramification

The distinction between tame and wild ramification is fundamental:

Tame ramification (char⁑k=0\operatorname{char} k = 0, or char⁑k=p>0\operatorname{char} k = p > 0 with p∀ePp \nmid e_P): the local picture is t↦tet \mapsto t^e, and the different exponent is dP=ePβˆ’1d_P = e_P - 1. The ramification is "polynomial" and well-behaved.

Wild ramification (char⁑k=p>0\operatorname{char} k = p > 0 with p∣ePp \mid e_P): the different exponent satisfies dPβ‰₯ePd_P \geq e_P, and higher ramification groups are nontrivial. The local structure is more intricate: it involves Artin-Schreier extensions in the pp-part.

Over C\mathbb{C}, all ramification is tame. Over fields of characteristic pp, wild ramification arises frequently and introduces subtle phenomena.

ExampleWild ramification: Artin--Schreier cover

Let kk have characteristic p>0p > 0. The Artin--Schreier cover f:Cβ†’P1f: C \to \mathbb{P}^1 is defined by ypβˆ’y=xβˆ’my^p - y = x^{-m} for gcd⁑(m,p)=1\gcd(m, p) = 1.

This is a degree-pp cyclic cover (Galois group Z/p\mathbb{Z}/p), totally ramified above x=∞x = \infty with ramification index e=pe = p. The different exponent at the unique point above ∞\infty is d=(pβˆ’1)(m+1)d = (p-1)(m+1), which is strictly greater than eβˆ’1=pβˆ’1e - 1 = p - 1 when mβ‰₯2m \geq 2.

For p=2p = 2, m=1m = 1: the curve y2βˆ’y=1/xy^2 - y = 1/x over F2\mathbb{F}_2. Here e=2e = 2, d=1β‹…2=2>1=eβˆ’1d = 1 \cdot 2 = 2 > 1 = e - 1. Wild!

For p=3p = 3, m=2m = 2: the curve y3βˆ’y=xβˆ’2y^3 - y = x^{-2} over F3\mathbb{F}_3. Here e=3e = 3, d=2β‹…3=6>2=eβˆ’1d = 2 \cdot 3 = 6 > 2 = e - 1. The genus of CC is larger than it would be with tame ramification.


Ramification and differentials

Definition4.16aPullback of differentials and the ramification divisor

For a separable morphism f:C→Df: C \to D, the pullback of the sheaf of differentials gives a map of line bundles:

fβˆ—Ξ©D1β†’Ξ©C1.f^* \Omega^1_D \to \Omega^1_C.

If tt is a local coordinate at Q∈DQ \in D and ss is a local coordinate at P∈CP \in C with f(P)=Qf(P) = Q, then fβˆ—t=uβ‹…sePf^* t = u \cdot s^{e_P} for a unit u∈OC,Pβˆ—u \in \mathcal{O}_{C,P}^*. Differentiating:

fβˆ—(dt)=d(uβ‹…seP)=(uβ€²seP+ePusePβˆ’1) ds=sePβˆ’1(uβ€²s+ePu) ds.f^*(dt) = d(u \cdot s^{e_P}) = (u' s^{e_P} + e_P u s^{e_P - 1})\, ds = s^{e_P - 1}(u' s + e_P u)\, ds.

In the tame case (char⁑k∀eP\operatorname{char} k \nmid e_P), the factor ePue_P u is a unit, so fβˆ—(dt)f^*(dt) vanishes to order exactly ePβˆ’1e_P - 1 at PP. Thus:

div⁑(fβˆ—Ο‰)=fβˆ—div⁑(Ο‰)+R\operatorname{div}(f^* \omega) = f^* \operatorname{div}(\omega) + R

for any nonzero rational differential Ο‰\omega on DD, where RR is the ramification divisor.

This gives the sheaf-theoretic relation: Ξ©C1β‰…fβˆ—Ξ©D1βŠ—OC(R)\Omega^1_C \cong f^* \Omega^1_D \otimes \mathcal{O}_C(R), or equivalently KC∼fβˆ—KD+RK_C \sim f^* K_D + R.


Hurwitz formula

Theorem4.12Hurwitz formula

Let f:C→Df: C \to D be a finite separable morphism of smooth projective curves of genera gCg_C and gDg_D, respectively. Then

2gCβˆ’2=deg⁑(f)β‹…(2gDβˆ’2)+deg⁑R2g_C - 2 = \deg(f) \cdot (2g_D - 2) + \deg R

where R=βˆ‘P∈C(ePβˆ’1)β‹…PR = \sum_{P \in C} (e_P - 1) \cdot P is the ramification divisor (using dPd_P instead of ePβˆ’1e_P - 1 in the wild case).

Equivalently, taking degrees in KC∼fβˆ—KD+RK_C \sim f^* K_D + R:

deg⁑KC=deg⁑fβ‹…deg⁑KD+deg⁑R.\deg K_C = \deg f \cdot \deg K_D + \deg R.

Since deg⁑K=2gβˆ’2\deg K = 2g - 2 for a curve of genus gg, the formula follows.

ExampleDouble cover of β„™^1 (hyperelliptic)

Let f:C→P1f: C \to \mathbb{P}^1 be a degree-2 cover with gD=0g_D = 0. The Hurwitz formula gives:

2gCβˆ’2=2(0βˆ’2)+deg⁑R=βˆ’4+deg⁑R.2g_C - 2 = 2(0 - 2) + \deg R = -4 + \deg R.

Each ramification point has eP=2e_P = 2 (totally ramified), so contributes ePβˆ’1=1e_P - 1 = 1 to RR. If there are rr ramification points, deg⁑R=r\deg R = r, and:

gC=rβˆ’22.g_C = \frac{r - 2}{2}.

So rr must be even, and:

  • r=2r = 2: gC=0g_C = 0 (an isomorphism after base change).
  • r=4r = 4: gC=1g_C = 1 (elliptic curve).
  • r=6r = 6: gC=2g_C = 2 (genus 2, all genus-2 curves are hyperelliptic).
  • r=2g+2r = 2g + 2: gC=gg_C = g (general hyperelliptic curve).
ExampleWeierstrass points of hyperelliptic curves

For a hyperelliptic curve C:y2=f(x)C: y^2 = f(x) where ff has degree 2g+22g + 2 with distinct roots Ξ±1,…,Ξ±2g+2\alpha_1, \ldots, \alpha_{2g+2}, the map Ο€:Cβ†’P1\pi: C \to \mathbb{P}^1, (x,y)↦x(x,y) \mapsto x has degree 2.

The ramification points are exactly the Weierstrass points: the 2g+22g + 2 points (Ξ±i,0)(\alpha_i, 0) where f(Ξ±i)=0f(\alpha_i) = 0. At each of these, the fiber is a single point with e=2e = 2.

When deg⁑f=2g+1\deg f = 2g + 1 instead (one root "at infinity"), the point at infinity is also a ramification point, giving r=2g+2r = 2g + 2 total. The Hurwitz formula still gives gC=gg_C = g.

For g=2g = 2, C:y2=x5βˆ’1C: y^2 = x^5 - 1: the 6 ramification points are (e2Ο€ik/5,0)(e^{2\pi i k/5}, 0) for k=0,…,4k = 0,\ldots,4 plus the point at infinity. The 6 Weierstrass points are the fixed points of the hyperelliptic involution (x,y)↦(x,βˆ’y)(x,y) \mapsto (x,-y).

ExampleIsogeny between elliptic curves

Let f:E1β†’E2f: E_1 \to E_2 be an isogeny of elliptic curves (both genus 1). The Hurwitz formula gives:

0=deg⁑(f)β‹…0+deg⁑R0 = \deg(f) \cdot 0 + \deg R

so deg⁑R=0\deg R = 0, meaning ff is unramified everywhere. This confirms the general fact: a homomorphism of abelian varieties is etale (unramified) if and only if it is an isogeny, which is automatic for isogenies between elliptic curves in characteristic 0 (or when the degree is coprime to the characteristic).


Cyclic covers and explicit ramification

ExampleCyclic covers of β„™^1

A cyclic cover of degree nn is f:Cβ†’P1f: C \to \mathbb{P}^1 given by yn=∏i=1r(xβˆ’Ξ±i)aiy^n = \prod_{i=1}^r (x - \alpha_i)^{a_i} where 0<ai<n0 < a_i < n and gcd⁑(ai,n)=1\gcd(a_i, n) = 1 ensures the cover is totally ramified above Ξ±i\alpha_i.

At each ramification point above Ξ±i\alpha_i, the ramification index is e=ne = n and (in the tame case, char⁑k∀n\operatorname{char} k \nmid n) the contribution to RR is nβˆ’1n - 1.

By Hurwitz: 2gCβˆ’2=n(2β‹…0βˆ’2)+r(nβˆ’1)=βˆ’2n+r(nβˆ’1)2g_C - 2 = n(2 \cdot 0 - 2) + r(n-1) = -2n + r(n-1), giving:

gC=(nβˆ’1)(rβˆ’2)2(ifΒ allΒ ai=1).g_C = \frac{(n-1)(r-2)}{2} \quad \text{(if all } a_i = 1 \text{)}.

For n=3n = 3, r=4r = 4 (four branch points): gC=2β‹…22=2g_C = \frac{2 \cdot 2}{2} = 2.

For n=4n = 4, r=3r = 3: gC=3β‹…12g_C = \frac{3 \cdot 1}{2}, which is not an integer! This means the ramification pattern with all ai=1a_i = 1 and r=3r = 3 is not compatible with n=4n = 4 unless we include the point at infinity, making r=4r = 4 and gC=3g_C = 3.

ExampleFermat curve projection

The Fermat curve Fn:xn+yn=znF_n: x^n + y^n = z^n in P2\mathbb{P}^2 is smooth of degree nn, so g=(nβˆ’1)(nβˆ’2)2g = \frac{(n-1)(n-2)}{2}.

The projection Ο€:Fnβ†’P1\pi: F_n \to \mathbb{P}^1, [x:y:z]↦[x:z][x:y:z] \mapsto [x:z] (i.e., t=x/zt = x/z) has degree nn. Setting s=y/zs = y/z, we have sn=1βˆ’tns^n = 1 - t^n.

The ramification occurs where dsdt\frac{ds}{dt} is undefined, i.e., where nsnβˆ’1dsdt=βˆ’ntnβˆ’1ns^{n-1} \frac{ds}{dt} = -nt^{n-1}, so dsdt=βˆ’tnβˆ’1/snβˆ’1\frac{ds}{dt} = -t^{n-1}/s^{n-1}, which blows up when s=0s = 0. This happens at t=ΞΆt = \zeta where ΞΆn=1\zeta^n = 1 (the nn-th roots of unity).

Over each branch point t=ΞΆt = \zeta, the fiber is a single point (x,y,z)=[ΞΆ:0:1](x,y,z) = [\zeta : 0 : 1] with e=ne = n. There are nn such branch points.

Hurwitz check: 2gβˆ’2=nβ‹…(βˆ’2)+n(nβˆ’1)2g - 2 = n \cdot (-2) + n(n-1) gives 2gβˆ’2=n2βˆ’3n=(nβˆ’1)(nβˆ’2)βˆ’22g - 2 = n^2 - 3n = (n-1)(n-2) - 2, which is g=(nβˆ’1)(nβˆ’2)2g = \frac{(n-1)(n-2)}{2}, matching the genus formula for plane curves.


Branch locus and monodromy

Definition4.16bBranch locus and monodromy representation

For a finite morphism f:Cβ†’Df: C \to D of degree nn over C\mathbb{C} (or more generally in the etale topology), the restriction f:Cβˆ–fβˆ’1(B)β†’Dβˆ–Bf: C \setminus f^{-1}(B) \to D \setminus B is an unramified covering space, where BβŠ‚DB \subset D is the branch locus.

The monodromy representation is the homomorphism

ρ:Ο€1(Dβˆ–B,Q0)β†’Sn\rho: \pi_1(D \setminus B, Q_0) \to S_n

that records how the nn sheets of the cover are permuted as we loop around branch points. The image Im⁑(ρ)\operatorname{Im}(\rho) is a transitive subgroup of SnS_n (since CC is connected), and the cover is Galois if and only if the image is a transitive subgroup acting regularly (equivalently, ∣Im⁑(ρ)∣=n|\operatorname{Im}(\rho)| = n).

The ramification index ePe_P at a point PP above a branch point QQ equals the length of the cycle in ρ(γQ)\rho(\gamma_Q) that contains the sheet corresponding to PP, where γQ\gamma_Q is a small loop around QQ.

ExampleMonodromy of double covers

For a degree-2 cover f:Cβ†’P1f: C \to \mathbb{P}^1 branched at {p1,…,p2g+2}\{p_1, \ldots, p_{2g+2}\}:

The monodromy group is S2=Z/2S_2 = \mathbb{Z}/2. Each branch point contributes a transposition (1 2)(1\,2). The monodromy representation is:

ρ:Ο€1(P1βˆ–{p1,…,p2g+2})β†’Z/2,Ξ³pi↦(1 2).\rho: \pi_1(\mathbb{P}^1 \setminus \{p_1, \ldots, p_{2g+2}\}) \to \mathbb{Z}/2, \quad \gamma_{p_i} \mapsto (1\,2).

The fundamental group relation Ξ³p1β‹―Ξ³p2g+2=1\gamma_{p_1} \cdots \gamma_{p_{2g+2}} = 1 requires (1 2)2g+2=id(1\,2)^{2g+2} = \mathrm{id}, which holds since 2g+22g+2 is even.

The cover is always Galois (since S2S_2 is abelian), with Galois group Z/2\mathbb{Z}/2. The nontrivial automorphism is the hyperelliptic involution (x,y)↦(x,βˆ’y)(x,y) \mapsto (x,-y).

ExampleMonodromy of a cubic cover

Consider f:Cβ†’P1f: C \to \mathbb{P}^1 given by y3=x(xβˆ’1)y^3 = x(x-1) (a degree-3 cyclic cover). The branch points are {0,1,∞}\{0, 1, \infty\}, and the monodromy representation maps into S3S_3:

ρ(Ξ³0)=(1 2 3),ρ(Ξ³1)=(1 2 3),ρ(γ∞)=(1 2 3)βˆ’1β‹…(1 2 3)βˆ’1=(1 2 3).\rho(\gamma_0) = (1\,2\,3), \quad \rho(\gamma_1) = (1\,2\,3), \quad \rho(\gamma_\infty) = (1\,2\,3)^{-1} \cdot (1\,2\,3)^{-1} = (1\,2\,3).

Wait -- the relation is Ξ³0Ξ³1γ∞=1\gamma_0 \gamma_1 \gamma_\infty = 1, so ρ(γ∞)=(1 2 3)βˆ’2=(1 2 3)\rho(\gamma_\infty) = (1\,2\,3)^{-2} = (1\,2\,3). The image is Z/3βŠ‚S3\mathbb{Z}/3 \subset S_3, confirming the cover is Galois with group Z/3\mathbb{Z}/3.

Hurwitz: 2gβˆ’2=3(βˆ’2)+3β‹…2=02g - 2 = 3(-2) + 3 \cdot 2 = 0, so g=1g = 1. The curve CC is an elliptic curve.


Belyi maps

ExampleBelyi maps and dessins d'enfants

A Belyi map is a finite morphism f:Cβ†’P1f: C \to \mathbb{P}^1 unramified outside {0,1,∞}\{0, 1, \infty\}.

Belyi's theorem: A smooth projective curve CC over C\mathbb{C} admits a Belyi map if and only if CC is defined over Qβ€Ύ\overline{\mathbb{Q}}.

Example 1: The identity id:P1β†’P1\mathrm{id}: \mathbb{P}^1 \to \mathbb{P}^1 is the simplest Belyi map (degree 1, no ramification).

Example 2: f(t)=4t(1βˆ’t)f(t) = 4t(1-t) is a degree-2 Belyi map. Branch points: fβ€²(t)=4βˆ’8t=0f'(t) = 4 - 8t = 0 at t=1/2t = 1/2, so f(1/2)=1f(1/2) = 1. The map is ramified at t=1/2t = 1/2 (above 11) and at t=∞t = \infty (above ∞\infty), with e=2e = 2 at each. Unramified outside {0,1,∞}\{0, 1, \infty\} since the only critical value is 11 (and ∞\infty).

Example 3: The Belyi map f(t)=(t2βˆ’t+1)3t2(tβˆ’1)2f(t) = \frac{(t^2 - t + 1)^3}{t^2(t-1)^2} gives a degree-6 cover of P1\mathbb{P}^1 by P1\mathbb{P}^1 (genus 0). The ramification type over 00 is [2,2,2][2, 2, 2], over 11 is [3,3][3, 3], and over ∞\infty is [2,2,2][2, 2, 2]. This corresponds to the dessin of a tetrahedron.

ExampleBelyi map from an elliptic curve

The elliptic curve E:y2=x(xβˆ’1)(xβˆ’Ξ»)E: y^2 = x(x-1)(x-\lambda) has a natural degree-2 map Ο€:Eβ†’P1\pi: E \to \mathbb{P}^1 via (x,y)↦x(x,y) \mapsto x, branched at {0,1,Ξ»,∞}\{0, 1, \lambda, \infty\}.

To make a Belyi map, we compose with a rational function g:P1β†’P1g: \mathbb{P}^1 \to \mathbb{P}^1 that sends {0,1,Ξ»,∞}\{0, 1, \lambda, \infty\} into {0,1,∞}\{0, 1, \infty\}. For instance, if Ξ»=βˆ’1\lambda = -1, we can use g(x)=x2g(x) = x^2, which sends {0,1,βˆ’1,∞}β†’{0,1,1,∞}\{0, 1, -1, \infty\} \to \{0, 1, 1, \infty\}.

The composition gβˆ˜Ο€:Eβ†’P1g \circ \pi: E \to \mathbb{P}^1 is then a degree-4 Belyi map on the curve y2=x(xβˆ’1)(x+1)=x3βˆ’xy^2 = x(x-1)(x+1) = x^3 - x.


Number field analogy

RemarkRamification in number fields

The analogy between function fields and number fields is one of the deepest themes in arithmetic geometry. For a number field extension L/KL/K (e.g., K=QK = \mathbb{Q}), the ring of integers OL\mathcal{O}_L is a "curve" and primes p\mathfrak{p} of KK play the role of points of DD:

A prime p\mathfrak{p} of KK factors in LL as pOL=P1e1β‹―Pgeg\mathfrak{p} \mathcal{O}_L = \mathfrak{P}_1^{e_1} \cdots \mathfrak{P}_g^{e_g} with βˆ‘eifi=[L:K]\sum e_i f_i = [L:K] where fi=[OL/Pi:OK/p]f_i = [\mathcal{O}_L/\mathfrak{P}_i : \mathcal{O}_K/\mathfrak{p}]. Then:

  • Split: ei=fi=1e_i = f_i = 1 for all ii, and g=[L:K]g = [L:K]. Analogous to an unramified point with deg⁑f\deg f distinct preimages.
  • Inert: g=1g = 1, e=1e = 1, f=[L:K]f = [L:K]. One preimage of higher residue degree.
  • Ramified: some eiβ‰₯2e_i \geq 2. Exactly the branch points of the "cover."

For Q(d)/Q\mathbb{Q}(\sqrt{d})/\mathbb{Q}: a prime pp ramifies iff pβˆ£Ξ”p \mid \Delta (discriminant). The branch locus is {p:p∣2d}\{p : p \mid 2d\}, a finite set -- just as for morphisms of curves.

ExampleGaussian integers as a double cover

The extension Z[i]/Z\mathbb{Z}[i]/\mathbb{Z} (i.e., Q(i)/Q\mathbb{Q}(i)/\mathbb{Q}) is a "degree-2 cover" of Spec⁑Z\operatorname{Spec} \mathbb{Z}:

  • p=2p = 2: (1+i)2=2i(1+i)^2 = 2i, so 2=βˆ’i(1+i)22 = -i(1+i)^2. Ramified (e=2e = 2). The prime 22 is the unique branch point.
  • p≑1(mod4)p \equiv 1 \pmod{4}: p=ppΛ‰p = \mathfrak{p} \bar{\mathfrak{p}} splits into two primes. Split (unramified, two preimages).
  • p≑3(mod4)p \equiv 3 \pmod{4}: pp remains prime in Z[i]\mathbb{Z}[i]. Inert (unramified, one preimage of degree 2).

This mirrors a double cover of curves: branch points (ramified primes), unramified points splitting into 2 preimages, and (unlike the geometric case) inert primes where the residue field doubles.


Ramification of specific families

ExampleHyperelliptic curve of genus 3

Let C:y2=(xβˆ’1)(xβˆ’2)(xβˆ’3)(xβˆ’4)(xβˆ’5)(xβˆ’6)(xβˆ’7)(xβˆ’8)C: y^2 = (x - 1)(x - 2)(x - 3)(x - 4)(x - 5)(x - 6)(x - 7)(x - 8) over C\mathbb{C}.

The projection Ο€(x,y)=x\pi(x,y) = x is a degree-2 map to P1\mathbb{P}^1 with r=8r = 8 branch points (the roots of the polynomial, all with e=2e = 2). By Hurwitz: g=(8βˆ’2)/2=3g = (8 - 2)/2 = 3.

The ramification divisor is R=βˆ‘i=18PiR = \sum_{i=1}^{8} P_i where Pi=(i,0)P_i = (i, 0), with deg⁑R=8\deg R = 8.

The canonical divisor is KC=Ο€βˆ—KP1+RβˆΌΟ€βˆ—(βˆ’2[∞])+RK_C = \pi^* K_{\mathbb{P}^1} + R \sim \pi^*(-2[\infty]) + R. Since deg⁑KC=2gβˆ’2=4\deg K_C = 2g - 2 = 4 and deg⁑R=8\deg R = 8, degβ‘Ο€βˆ—KP1=2β‹…(βˆ’2)=βˆ’4\deg \pi^* K_{\mathbb{P}^1} = 2 \cdot (-2) = -4, and indeed βˆ’4+8=4-4 + 8 = 4.

ExampleThe Klein quartic

The Klein quartic C:x3y+y3z+z3x=0C: x^3 y + y^3 z + z^3 x = 0 in P2\mathbb{P}^2 is a smooth plane curve of degree 4, so g=3g = 3.

It has the remarkable automorphism group Aut⁑(C)β‰…PSL⁑2(F7)\operatorname{Aut}(C) \cong \operatorname{PSL}_2(\mathbb{F}_7) of order 168, achieving the Hurwitz bound ∣Aut⁑(C)βˆ£β‰€84(gβˆ’1)=168|\operatorname{Aut}(C)| \leq 84(g-1) = 168.

The quotient map f:Cβ†’C/Aut⁑(C)β‰…P1f: C \to C/\operatorname{Aut}(C) \cong \mathbb{P}^1 has degree 168. By Hurwitz:

2(3)βˆ’2=168(0βˆ’2)+deg⁑Rβ€…β€ŠβŸΉβ€…β€Šdeg⁑R=4+336=340.2(3) - 2 = 168(0 - 2) + \deg R \implies \deg R = 4 + 336 = 340.

The ramification has three orbits of branch points in P1\mathbb{P}^1:

  • Above one branch point: 24 points with e=7e = 7, contributing 24β‹…6=14424 \cdot 6 = 144.
  • Above another: 42 points with e=4e = 4, contributing 42β‹…3=12642 \cdot 3 = 126.
  • Above the third: 56 points with e=3e = 3, contributing 56β‹…2=11256 \cdot 2 = 112. -- but wait, let us recheck: we need 24β‹…7=16824 \cdot 7 = 168, 42β‹…4=16842 \cdot 4 = 168, 56β‹…3=16856 \cdot 3 = 168. Contribution to deg⁑R\deg R: 24β‹…6+42β‹…3+56β‹…2=144+126+112=38224 \cdot 6 + 42 \cdot 3 + 56 \cdot 2 = 144 + 126 + 112 = 382. But we need deg⁑R=340\deg R = 340.

The correct ramification data is: above three branch points, with orbit sizes 24,84,5624, 84, 56 and ramification indices 7,2,37, 2, 3:

  • 24 points with e=7e = 7: contribution 24β‹…6=14424 \cdot 6 = 144.
  • 84 points with e=2e = 2: contribution 84β‹…1=8484 \cdot 1 = 84.
  • 56 points with e=3e = 3: contribution 56β‹…2=11256 \cdot 2 = 112.

Total: 144+84+112=340144 + 84 + 112 = 340. Check: 24β‹…7=16824 \cdot 7 = 168, 84β‹…2=16884 \cdot 2 = 168, 56β‹…3=16856 \cdot 3 = 168. Correct!


Purity and the branch locus

TheoremPurity of the branch locus

Let f:Cβ†’Df: C \to D be a finite morphism of smooth curves. If ff is unramified at every point in fβˆ’1(Q)f^{-1}(Q) for some Q∈DQ \in D, then ff is etale in a neighborhood of fβˆ’1(Q)f^{-1}(Q).

More generally (Zariski--Nagata purity): for a finite morphism f:X→Yf: X \to Y of smooth varieties, the branch locus (the set of points of YY over which ff is not etale) is a pure codimension-1 subvariety. For curves, this means the branch locus is a finite set of closed points.


Riemann--Hurwitz and topology

RemarkTopological viewpoint over β„‚

Over C\mathbb{C}, the Hurwitz formula is a consequence of the multiplicativity of the Euler characteristic for covering spaces, corrected by ramification.

For f:C→Df: C \to D of degree nn, removing the branch locus and its preimage gives an honest nn-sheeted covering f0:C0→D0f_0: C_0 \to D_0. The Euler characteristics satisfy χ(C0)=n⋅χ(D0)\chi(C_0) = n \cdot \chi(D_0).

Filling back in the removed points: each branch point QiQ_i adds 11 to Ο‡(D0)\chi(D_0), and the points above QiQ_i add ∣fβˆ’1(Qi)∣|f^{-1}(Q_i)| to Ο‡(C0)\chi(C_0). Since ∣fβˆ’1(Qi)∣=βˆ‘P↦Qi1=nβˆ’βˆ‘P↦Qi(ePβˆ’1)|f^{-1}(Q_i)| = \sum_{P \mapsto Q_i} 1 = n - \sum_{P \mapsto Q_i}(e_P - 1), we get:

Ο‡(C)=nβ‹…Ο‡(D)βˆ’βˆ‘P∈C(ePβˆ’1)=nβ‹…Ο‡(D)βˆ’deg⁑R.\chi(C) = n \cdot \chi(D) - \sum_{P \in C}(e_P - 1) = n \cdot \chi(D) - \deg R.

Since Ο‡=2βˆ’2g\chi = 2 - 2g for a compact Riemann surface, this is the Hurwitz formula.


Further examples

ExampleTriple cover producing genus 2

The curve C:y3=x(xβˆ’1)C: y^3 = x(x-1) is a degree-3 cyclic cover of P1\mathbb{P}^1 via (x,y)↦x(x,y) \mapsto x.

Branch points: x=0x = 0 and x=1x = 1 (where y3=0y^3 = 0, a single root with e=3e = 3), and x=∞x = \infty (where the equation becomes y3=x2y^3 = x^2 in suitable coordinates, again e=3e = 3).

So r=3r = 3 branch points, each totally ramified with e=3e = 3. By Hurwitz:

2gβˆ’2=3(βˆ’2)+3(3βˆ’1)=βˆ’6+6=0.2g - 2 = 3(-2) + 3(3-1) = -6 + 6 = 0.

Thus g=1g = 1: this is an elliptic curve, not genus 2!

To get genus 2, we need more ramification: C:y3=x(xβˆ’1)(xβˆ’2)(xβˆ’3)C: y^3 = x(x-1)(x-2)(x-3). Here the branch points are {0,1,2,3}\{0, 1, 2, 3\}, but each root appears with exponent 1, and since gcd⁑(1,3)=1\gcd(1,3) = 1 each is totally ramified. The exponent sum is 4≑1(mod3)4 \equiv 1 \pmod{3}, so ∞\infty is also a branch point (since the total degree 4≑̸0(mod3)4 \not\equiv 0 \pmod{3}), but there we need a separate analysis. Ultimately this gives a curve of genus g=(3βˆ’1)(5βˆ’2)2=3g = \frac{(3-1)(5-2)}{2} = 3 (with 5 branch points).

ExampleMap from a smooth quartic to β„™^1

Let CC be a smooth plane quartic (g=3g = 3). Projecting from a point P∈CP \in C gives Ο€P:Cβ†’P1\pi_P: C \to \mathbb{P}^1 of degree 3 (a line through PP meets CC in 44 points, one of which is PP, leaving 33 others).

By Hurwitz: 2(3)βˆ’2=3(βˆ’2)+deg⁑R2(3) - 2 = 3(-2) + \deg R gives deg⁑R=10\deg R = 10. So there are 1010 ramification points (each with e=2e = 2, generically), corresponding to the lines through PP that are tangent to CC at some other point.

Geometrically: 10=4β‹…3βˆ’2=10 = 4 \cdot 3 - 2 = the number of tangent lines from PP to CC other than the tangent at PP itself (by the class formula for plane curves).


Summary

RemarkKey takeaways

The essential facts about ramification for curves:

  1. Ramification index eP=vP(fβˆ—t)e_P = v_P(f^*t) measures the local branching order.
  2. Hurwitz formula: 2gCβˆ’2=deg⁑(f)(2gDβˆ’2)+deg⁑R2g_C - 2 = \deg(f)(2g_D - 2) + \deg R relates genera, degree, and ramification.
  3. Tame vs wild: in characteristic pp, wild ramification (p∣ePp \mid e_P) gives a larger different, requiring more care.
  4. Differentials detect ramification: Ξ©C1β‰…fβˆ—Ξ©D1βŠ—OC(R)\Omega^1_C \cong f^*\Omega^1_D \otimes \mathcal{O}_C(R).
  5. Monodromy encodes the global structure: the cover is determined (up to isomorphism) by its monodromy representation Ο€1(Dβˆ–B)β†’Sn\pi_1(D \setminus B) \to S_n.
  6. Number field analogy: primes splitting, remaining inert, or ramifying in extensions mirror the behavior of fibers of morphisms of curves.