Computations of Cannonical Divisors and Sheaves

There is an intrinsic classification of (smooth) curves by cannonical divisor. In particular, if the anticannonical class is ample, the curve is P1\mathbb P^1; if neither the anticannonical class nor the cannonical class is ample the curve is an elliptic curve; and if the cannonical class is ample the curve is “of general type”. Note that these conditions correspond to the genus of the curve being 00, 11, and >1>1, respectively; furthermore, they are equal to the possible values for the Kodaira dimension. This classification for curves is in direct analogy with the classification of surfaces, which I hope to write my thesis on. Here we compute some examples of the cannonical divisor to get a feel for it.

The cannonical divisor of P1\mathbb P^1.

I’ll type this up eventually.

The cannonical divisor of the Fermat cubic.

Let X=V(x3+y3+z3)P2X = V(x^3 + y^3 + z^3)\subset \mathbb P^2, and consider the two open affines U0=U(z)U_0 = U(z) and U1=U(y)U_1 = U(y). We work first in U0U_0, letting u=xzu = \frac xz and v=yzv = \frac yz. Then U0Speck[u,v]/(u3+v3+1)U_0 \cong \spec k[u, v] / (u^3+v^3 + 1), so that

ωX(U1)ΩX(U1)((k[u,v]/(u3+v3+1))(dudv))/(u2du+v2dv).\omega_X(U_1) \cong \Omega_X(U_1) \cong ((k[u, v] / (u^3+v^3 +1))(du \oplus dv))/(u^2du + v^2dv).

We consider the section of (ωXK(X))(U)(\omega_X\otimes K(X))(U) (rational sections of ωX\omega_X, where K(X)K(X) is the function field of XX) given by α=duv2=dvu2\alpha = \frac{du}{v^2} = -\frac{dv}{u^2}. Restricting to U(u)U(u), we have that dvu2-\frac{dv}{u^2} is regular and nonvanishing; if it vanished at a point pp (that is, if it lived in the image of the maximal ideal under the isomorphism OpωX,p\mathcal O_p\cong \omega_{X,p}), then we would have dv=0dv=0 which implies (by u2du=v2dv=0u^2du=-v^2dv=0) that dudu is also in the image of the maximal ideal. Since dudu and dvdv generate the sheaf of differentials, this would mean that there is a point pp where the entire sheaf of differentials lives in the image of the maximal ideal of OX,p\mathcal O_{X, p}, a contradiction. Restricting to U(v)U(v) we can write α=duv2\alpha = \frac{du}{v^2} and apply a symmetric argument, giving that α\alpha is a regular differential - a section of ωX\omega_X - on U0U_0.

Now on U1U_1 let t=xyt=\frac xy and s=zys=\frac zy, and look at α=dts2=dst2\alpha' = -\frac{dt}{s^2} = \frac{ds}{t^2}. By the same argument (thanks to the symmetry of the Fermat cubic!) we have that α\alpha' is regular on this open set. Furthermore, if we restrict further to U(s)U(s) we get isomorphisms U0V(s)U1V(v)U_0\setminus V(s)\to U_1\setminus V(v) given by the map on rings

f:k[t,s,s1]/(t3+s3+1)k[u,v,v1]/(u3+v3+1),f:k[t, s, s^{-1}]/(t^3+s^3+1) \to k[u, v, v^{-1}]/(u^3+v^3+1),

s1v,tuvs\mapsto \frac1v, t\mapsto \frac uv

This induces a map f~\tilde f on differentials over these open sets, given by

ds1v2dv,dtvduudvv2ds \mapsto -\frac1{v^2}dv, dt\mapsto \frac{vdu - udv}{v^2}

Note that this map (extended to a map on rational sections) takes α=dst2\alpha'= \frac{ds}{t^2} to f(1t2)f~(ds)=v2u21v2dv=dvu2=αf\left(\frac 1{t^2}\right)\tilde f(ds) = -\frac{v^2}{u^2}\frac 1{v^2}dv = -\frac{dv}{u^2} = \alpha. Thus α\alpha and α\alpha' glue on the intersection U1U0=U1V(v)U_1\cap U_0 = U_1\setminus V(v) to obtain a global differential form (note that U1U_1 and U0U_0 cover XX, because if XX and YY vanish then ZZ must likewise vanish).

This differential form vanishes nowhere, and thus gives an isomorphism ωXOX\omega_X\cong \mathcal O_X globally.