MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let $C$ be a complex algebraic curve. It is well known that if $L$ is a special divisor on $C$, i.e., $h^0(L) > 0$ and $h^1(L) > 0$, then $$ h^0 (L) \le \frac{1}{2} \deg L + 1. $$ Assume that $L$ is not trivial and $L \ne K_C$. The equality holds if and only if $C$ is hyperelliptic.

My question is that: if there is a nontrivial special divisor $L$ on $C$ such that $L \ne K_C$ $$ h^0 (L) = \frac{1}{2} \deg L $$ holds, then what conditions $C$ should satisfy? For example, $C$ could be hyperelliptic. Is there a complete classification?

share|cite|improve this question
This questions seems unclear at the moment. First of all, every curve has a special line bundle satisfying the equality $h^0(L) = \frac{1}{2}deg(L)+1$; namely the canonical bundle. Hyperelliptic curves are unique because they also possess a line bundle which satisfies the equality which is neither the trivial bundle nor the canonical bundle. So, is your question asking which curves possess some special line bundle satisfying $h^0(L) = \frac{1}{2}deg(L)$? Or, are you asking for a list of all line bundles on all curves which satisfy the equality, akin to Clifford's theorem? – Mike Skirvin Nov 19 '11 at 17:29
One more comment: If you are simply looking for curves which possess a line bundle satisfying the equality, then any curve possessing a $g^1_4$ obviously works. This includes all hyperelliptic curves, but also all curves of genus at most $6$. More generally, you might try checking when the Brill-Noether number associated to $d=deg(L)$ and $r=\frac{1}{2}deg(L)-1$ is non-negative. I haven't done the calculation, but it seems that using this strategy you might find that every curve possesses a line bundle satisfying your equality? – Mike Skirvin Nov 19 '11 at 17:37
Right. I should assume that $L$ is not canonical bundle. Thank you! – Tong Nov 19 '11 at 22:41
Have you looked at Arbarello-Cornalba-Griffiths-Harris book? I don't have it with me right now, but is the first place I'd look. – Felipe Voloch Nov 20 '11 at 0:03

Notice first of all that if $|L|$ has a fixed part, then applying Clifford's theorem to the moving part of $|L|$ one obtains immediately that the fixed part is a point, $C$ is hyperelliptic and $|L|$ is a multiple of the $g^1_2$. So we can assume that $|L|$ is free.
Another remark is that $K_C-L$ also satisfies $h^0(K_C-L)=\deg(K_C-L)/2$, so we can also assume $\deg L\le g(C)-1$.

There exists a more precise form of Clifford's theorem, sometimes called "Clifford+" that can be useful for answering your question. The statement is the following (cf. [Beauville, "L'application canonique pour les surfaces de type g\'en\eral", Invent. Math. 55, 121-140, (1979], Lemma 5.1):

Let $C$ be a smooth curve and $L$ a divisor of $C$ with $\deg L\le g(C)-1$. Then one of the following holds:

(i) $h^0(L)\le \deg L/3+4$;

(ii) there exists a a degree 2 map $f\colon C\to \Gamma$ such that $\Gamma$ is a smooth curve and the moving part of $|L|$ is equal to $f^*|\Delta|$ where $\Delta$ is a divisor of $\Gamma$. In this case $h^0(L)\le \deg L/2+1-g(\Gamma)$.

In your case you get immediately that for $h^0(L)\ge 13$ the map given by $|L|$ factors through a double cover $C\to \Gamma$, where $\Gamma$ is either rational or elliptic. If $\Gamma$ is rational, then $C$ is hyperelliptic and $L$ is multiple of the $g^1_2$, but this is not possible since we are assuming that $2h^0(L)=\deg L$. So $C$ is bielliptic and $|L|$ is pulled back from an elliptic curve.

The cases with $r:=h^0(L)-1\le 11$ I think should be done by hand. For $r=1$ any tetragonal curve will do, as pointed out in the comments. For $r>1$, the image $C'$ of $C$ via $|L|$ has degree $\ge r$, hence if $r>2$ then the map given by $|L|$ has degree $\le 2$. If the degree is 2 then $C'$ has degree $r+1$ hence $C'$ is elliptic (it cannot be rational, otherwise the system $|L|$ would not be complete) and we are in the previous situation. If the map given by $|L|$ is birational, then you can use Castelnuovo's bound to give an upper bound for $g(C)$ in terms of $r$. In case $r=2$ there's one more possibility: $C'$ is a conic and $|L|$ gives a map of degree 3. This corresponds to the case in which $C$ is a trigonal curve and $L$ is twice a $g^1_3$.

share|cite|improve this answer

I'm going to assume $L$ is base point free. I think it is clear how to change what I have written in the case where there are base points (numbers goes down by the degree of the base locus). In general, if $L$ is a line bundle of degree $d$ and $h^0(C,L)= r+1$, then the Clifford index of $L$, written Cliff(L) $= d-2r$. Cliffords theorem is Cliff(L) >= 0. Your line bundle satisfies Cliff(L) = 2. Two ways to achieve that if for the curve $C$ to have a $g^1_4$ or be a plane sextic. The Clifford index of a curve is the min{cliff(L)| $h^0(L)$ & $h^1(L)$ >=2}. For a general curve $C$, Cliff(C) is the floor of (g-1)/2. So for a general curve this can't be done. It is absolutely true that A-C-G-H will have more details and I believe a careful perusal will give the classification of all curves with Clifford Index 2. I think it is only the cases I mention.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.