Surfaces can be classified using Kodaira dimensions and birational invariants. Denote by $\kappa(S)$ the Kodaira dimension of a smooth projective surface $S$. For a smooth projective surface, an equivalent characterization of Kodaira dimension uses the pluricanonical genus $P_{12}=\dim H^2(S, 12K_S)$ and $K_S^2$. When $K_S^2=0$ and $K_S$ is nef, we see that $0\leq \kappa(S)\leq 1$.

Is there an easier way (which means not using the classification of surfaces of $\kappa=0$) to proof that that $P_{12}=1$ and $K_S$ is nef implies $\kappa(S)=0$ or that $P_{12}\geq 2$, $K_S^2=0$ if $K_S$ is nef and $\kappa(S)=1$.

  • $\begingroup$ The second question seems easier. The assertions $K$ nef and $K^2=0$ imply that if a multiple of $K$ moves, then it maps to a curve. Combining this with the inequality on the 12th plurigenus gives you the result. I am not sure how to proceed in the other case: you would need to know something similar to $K^2=0$ to proceed in a similar way. $\endgroup$ – M P Apr 29 '13 at 8:04
  • $\begingroup$ Thanks, MP. You are right, the second one is easier. I should ask the other way around. $\endgroup$ – Fei YE Apr 30 '13 at 9:30

All of the following is from Beauville's wonderful, short, dense book: "Complex Algebraic Surfaces." I really recommend it if you want to learn about the classification or even general techniques in surface theory.

For a minimal surface, $P_{12}=1$ alone is equivalent to $\kappa=0$. It probably is necessary to go through the classification to get this result. The number $12$ is not random: It arises because of the bi-elliptic surfaces, which are quotients of a product of two elliptic curves by a finite group. These finite groups all have elements whose order divides $12$, which is why the number $12$ appears.

The key step is that $P_{12}=0$ implies $S$ is ruled. This follows from two cases:

(1) Suppose $q=0$. Because $P_{12}=0$ we also have $P_2=0$. Castelnuovo's rationality criterion then applies to show $S$ is rational.

(2) Suppose $q\geq 1$. Because $P_{12}=0$ we also have $p_g=0$. One can show relatively easily using the Albanese fibration that the only possibility for $S$ not to be ruled is when $q=1$ and $b_2=2$. Then lots of work shows that if $S$ were not ruled we would have $S=(B\times F)/G$ for curves $B$ and $F$ and a finite group $G$ (and a number of other technical restrictions). By analyzing the canonical bundle on the resulting surface, one can show that $P_{12}$ is never zero. Thus we have a contradiction, so $S$ is ruled. Furthermore, $P_{12}=1$ if and only if both $B$ and $F$ are elliptic curves.

Once this difficult step is out of the way, we know that $\kappa=0$ implies $P_{12}=1$. Conversely, if $P_{12}=1$ then $S$ is non-ruled, and one can show $\chi(O_S)\geq 0$. Since $p_g=0$ or $1$, the list of possibilities for $q$ is finite. After analyzing the five cases (one turns out to be impossible), dealing with the most difficult ones by invoking part (2) from above, one can show that $P_{12}=1$ does in fact imply $\kappa=0$.

Finally, the remaining case is $P_{12}\geq 2$. In this case, it is simple to show that $\kappa=2$ if and only if $K^2>0$. The backward direction is an application of Riemann-Roch, and the forward direction is proven by contrapositive. If $K^2=0$, then the mobile part $M$ of $|nK|$ satisfies $M^2=0$ and hence $\kappa=1$.

Hope this helps. - Phil

  • $\begingroup$ Dear Philip, thanks for the answer. Maybe my question is not clear. The classification method is exactly what I want to avoid. I am looking for a direct way of proof. Thanks again. $\endgroup$ – Fei YE May 1 '13 at 5:05
  • $\begingroup$ One direct way to show $\kappa(S)=-\infty$ implies $S$ is ruled is to use Iitaka's Conjecture: If $S$ is a surface and $S\rightarrow B$ is a fibration with generic fiber $F$ then $\kappa(S)\geq\kappa(B)+\kappa(F)$. Applying this to the Albanese fibration solves case (2) above instantaneously, because it implies that $F$ must be rational. This applies the stronger assumption $P_n=0$ for all $n$ rather than $P_{12}=0$ though. I think there really will be no way to get the specific number $12$ without some classification. $\endgroup$ – Philip Engel May 1 '13 at 7:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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