Hurwitz Formula
The Hurwitz formula is the fundamental tool for computing the genus of a curve that admits a finite morphism to another curve. It relates the genera of the source and target curves via the degree of the map and a correction term measuring ramification. Together with Riemann--Roch, it forms the backbone of classical curve theory.
Ramification divisor
Let be a non-constant morphism of smooth projective curves over an algebraically closed field . The degree is the degree of the field extension .
For each point , the ramification index is the valuation , where is a uniformizer at . Equivalently, is determined by where is a unit in .
- is a ramification point if .
- is a branch point if some has .
- For every , the fundamental identity holds: .
The ramification divisor is .
The ramification at is tame if , and wild if . In characteristic , all ramification is tame. In characteristic , wild ramification introduces additional subtleties: the contribution of to the ramification divisor can exceed .
Statement of the Hurwitz formula
Let be a non-constant separable morphism of smooth projective curves of genera and over an algebraically closed field . Let and let be the ramification divisor. Then:
In the tame case (characteristic , or characteristic with for all ), we have , and the formula becomes:
The Hurwitz formula is equivalent to the relation of canonical divisors:
Taking degrees of both sides: , i.e., . The divisor-level statement is stronger: it says that the sheaf of differentials on is obtained from pulling back differentials on and adding zeros at ramification points.
Locally, if where is a uniformizer on and on , then . In the tame case (), this equals , which vanishes to order exactly at .
Wild ramification
Let be a non-constant separable morphism of smooth projective curves over an algebraically closed field of characteristic . Then:
where is the wild part of the different at :
- if (tame ramification),
- if (wild ramification).
More precisely, is determined by the higher ramification groups: if denotes the -th ramification group at (in lower numbering), then , which simplifies to .
Over a field of characteristic , the Artin--Schreier cover defines a degree- cover .
Consider with over . The map has degree , and:
- Over , there is a single point with and wild ramification. The different exponent is when computed carefully.
- All other fibers are unramified (for generic ).
By the Hurwitz formula: , giving , so when is chosen appropriately.
Compare this with the tame case: for in characteristic with , the genus formula yields a different answer because there is no wild contribution.
Applications: computing genera
A hyperelliptic curve of genus is a smooth projective curve admitting a degree- map . Applying Hurwitz with , :
Each ramification point has (since the degree is , and any ramification index must divide the degree), so at each ramification point. If there are ramification points:
Conclusion: A hyperelliptic curve of genus (in characteristic ) has exactly branch points in .
For example: genus (elliptic) has branch points, genus has , genus has .
Consider where is a squarefree polynomial of degree , with . The projection gives a degree- map .
Ramification occurs at the roots of (where , so there is a single point above each root) and possibly at :
- If is odd: there is one point above with , so ramification points.
- If is even: there are two points above with , so ramification points.
Hurwitz gives , so:
- odd: ,
- even: .
In both cases, . For instance: has genus ; also has genus .
Let where is squarefree, , and divides . The projection to gives a degree- map .
At each root , there is a unique point above with (when or handled carefully). The total ramification over a finite branch point contributes for points above it, where we need to track the local structure.
In the simplest case where and all ramification is total (single point above each with ): each branch point contributes to , and over the fiber is unramified. Then:
giving . More carefully:
when and . For , this recovers for even .
The Fermat curve in is a smooth curve of degree (when ).
Method 1 (degree-genus formula): For a smooth plane curve of degree :
Method 2 (Hurwitz formula): Consider the projection from to . For each value of , we solve .
- If : there are distinct solutions, so the fiber is unramified.
- If (the -th roots of unity): then , so there is a single point with .
- At (i.e., ): the equation becomes , giving distinct points (since ).
There are branch points (the -th roots of unity), each with a single point above having , contributing each. Hurwitz gives:
so , confirming the degree-genus formula.
Let be a degree- cover (in characteristic ). Then:
Possible ramification types over a branch point :
- (3): One point above with , contributing to .
- (2, 1): One point with and one with , contributing to .
If we have branch points of type and of type , then and:
Since is an integer, must be even. For genus : need , giving . For genus : need .
Concrete example: . This is a degree- cover of totally ramified over , , and , so , , giving . It is an elliptic curve.
An elliptic curve over (with ) is always a double cover of branched at points. In Weierstrass form :
- The map is a degree- map .
- Ramification at (where ) and at (after compactification).
- Hurwitz: , i.e., β.
The branch points determine up to isomorphism (modulo automorphisms of ). Since is -dimensional and we have points, the moduli space of elliptic curves is -dimensional, parametrized by the -invariant.
Constraints on existence of morphisms
Let be a non-constant separable morphism of smooth projective curves of genera and with . Then:
- , with equality if and only if is unramified.
- , with equality if and only if either (isomorphism) or and is unramified.
- If : , so . Since , this is always satisfiable.
- If : , so , with iff is unramified (an isogeny).
- If : , so , which gives a finite bound on the degree.
Hurwitz gives strong non-existence results:
A genus- curve cannot cover a genus- curve non-trivially. If with and , Hurwitz gives . Since , we need . So , i.e., is an isomorphism.
A genus- curve can cover a genus- curve only by a degree- map (isomorphism is impossible since genera differ) or... Hurwitz: , so . For : , so must be unramified of degree . Such maps exist (unramified double covers correspond to -torsion in , which is ).
No map from to a curve of genus . If , then . For : the right side is , contradiction.
Luroth's theorem
Let be an algebraically closed field, and let be a field with and . Then for some , i.e., is a purely transcendental extension of .
Geometric reformulation: If is a smooth projective curve admitting a non-constant map , then .
If has degree , then Hurwitz gives:
Since and , we need , hence , so and .
This is one of the simplest applications of the Hurwitz formula: the projective line can only map surjectively onto a curve of genus , which must again be the projective line. This fails in higher dimensions (Luroth's problem: not every unirational variety is rational).
de Franchis theorem
Let be a smooth projective curve of genus over an algebraically closed field (of characteristic ). Then for any smooth projective curve with , the set of non-constant morphisms is finite.
Moreover, the total number of non-constant morphisms from to all curves of genus is bounded by a constant depending only on .
The Hurwitz formula gives a bound on the degree: if has degree with , then . So the degree is bounded.
For a fixed degree , the map is determined by the field extension of degree . The finiteness of such extensions (for fixed source and bounded degree) requires deeper arguments beyond Hurwitz alone, involving rigidity of maps between hyperbolic curves.
The effective bound due to Howard and Sommese is: (though this is far from sharp in practice).
If , then Hurwitz gives for an unramified map, and for ramified maps , which forces . So the only maps from a genus- curve to another genus- curve are isomorphisms, and there are finitely many (in fact, the automorphism group of a genus- curve is finite, of order at most in characteristic ).
Unramified covers and the Euler characteristic
Over , the Hurwitz formula has a transparent topological interpretation. The Euler characteristic of a Riemann surface of genus is .
For a branched cover of degree , start with the naive expectation (which holds for unramified covers). Each ramification point "reduces" the Euler characteristic: instead of preimages of , we have only preimage contributing instead of to the fiber count. The correction is:
which is , recovering the Hurwitz formula.
An unramified double cover with gives , so .
- Double cover of genus : (an isogeny between elliptic curves).
- Double cover of genus : .
- Double cover of genus : .
Unramified double covers of are classified by , giving non-trivial covers (up to isomorphism as covers). For : there are unramified double covers, each a genus- curve.
Further applications
Let be a curve of genus admitting an involution (an automorphism of order , with ). The quotient is a smooth curve, and is a degree- map. Let be the genus of and let be the number of fixed points of (which are exactly the ramification points). By Hurwitz:
Since : , i.e., .
Special cases:
- (hyperelliptic involution): .
- : , so . For , (isogeny).
- : , a "bi-elliptic" situation when and .
Consider the tower with , . Applying Hurwitz twice:
Substituting: .
Alternatively, applying Hurwitz to of degree :
Comparing: . This reflects the multiplicativity of ramification: .
Belyi's theorem states that a smooth projective curve over admits a morphism branched over at most points (which may be taken to be ).
For such a Belyi map of degree branched over , the Hurwitz formula gives:
The ramification data over each branch point defines a partition of : if the fiber over has ramification indices , this gives a partition with contribution to .
If the three partitions have , , parts respectively, then , giving:
For genus Belyi maps: . For genus : .
Let be a finite group acting on a smooth projective curve over (with ). The quotient is a smooth curve, and is a Galois cover of degree .
For a point , the stabilizer at any has order , and . The Hurwitz formula becomes:
where for any above (well-defined since the cover is Galois).
This form is especially useful: the sum is taken over branch points of , and the contribution of each branch point depends only on the ramification index.
The Klein quartic is the curve in , which is smooth of degree and genus .
It has automorphism group of order , the maximum for genus (by the Hurwitz bound ).
The quotient has genus , and the Hurwitz formula gives:
So where . For : , giving . One checks that β.
So is branched at points with ramification indices . This is the origin of the famous triangle group.
The Hurwitz bound on automorphisms
Let be a smooth projective curve of genus over an algebraically closed field of characteristic . Then:
Let and with genus . The Hurwitz formula in the Galois form gives:
The key is to show that whenever this expression is positive.
If : the expression is , giving .
If : the expression equals , giving .
If with branch points: the expression is . For this to be positive with , we get , so . For with , the minimum positive value of occurs at , giving .
Hence , with equality only for the signature (as in the Klein quartic).
Genus of curves defined by equations
Let be a smooth curve of degree . Projecting from a point gives of degree (a generic line through meets in points).
For a general projection, the ramification occurs at the points where the tangent line to passes through . By a class computation (or Plucker formulas), the number of tangent lines from a general point to a smooth degree- curve is , and each tangent point has . So , and:
giving , the degree-genus formula.
A smooth complete intersection of hypersurfaces of degrees is a curve of degree and genus:
This follows from the adjunction formula rather than Hurwitz directly, but one can verify it via Hurwitz by projecting to and computing ramification.
Examples:
- Curve of type in : , . An elliptic curve!
- Curve of type in : , .
- Curve of type in : , .
Riemann--Hurwitz and differentials
The Hurwitz formula follows from the theory of differentials (Kahler differentials in the algebraic setting, or holomorphic -forms in the analytic setting).
Given , the pullback of differentials defines a map . If is a local coordinate on and is a local coordinate on with (tame case), then:
Since is a unit at (as in the tame case and ), the pullback differential vanishes to order exactly .
Therefore for any meromorphic differential on , giving at the level of divisor classes.
Summary
The Hurwitz formula is indispensable in algebraic geometry:
- Genus computation: determines the genus of curves defined by equations, branched covers, and quotients by group actions.
- Existence obstructions: constrains which maps between curves can exist, limiting both the degree and the ramification data.
- Automorphism bounds: leads to the bound on automorphism groups and the theory of Fuchsian groups.
- Moduli theory: the Hurwitz scheme parametrizing branched covers of with given ramification data is a fundamental object, connected to symmetric group representations and the braid group.
- Arithmetic applications: via Belyi's theorem, connects the absolute Galois group to combinatorial objects (dessins d'enfants) through the Hurwitz formula's constraints on ramification.