Let $S$ be a connected noetherian normal scheme of dimension 0 or 1 (i.e. $S$ is a connected Dedekind scheme).

Let $f:G\to G'$ be a morphism of semi-abelian schemes over $S$. In their book on Neron models, Bosch-Lutkebohmert-Raynaud define that $f$ is an isogeny if for all points $s\in S$ the base changed morphism $f_s:G_s\to G'_s$ is an isogeny (i.e. $f_s$ is finite and surjective on identity components).

Is $\deg(f_s)=\deg(f_{s'})$ for all $s,s'\in S$?

If this is the case, can one then define the degree of $f$ just by $\deg(f_s)$ for $s$ the generic point of $S$?

If the above questions have negative answers, can one answer them affirmatively for abelian schemes?


1 Answer 1


I am going to assume that $S$ is Noetherian. By limit arguments, one should always be able to reduce to that case. Also I am going to assume that your semi-Abelian schemes are smooth over $S$; I believe this is part of the usual definition.

In fact, $f$ should be flat, which implies constancy of the degree of $f_s$. The point is that both $G$ and $G'$ are smooth over $S$. Thus the fiber product, $G\times_S G'$, is flat over $G'$ of relative dimension $g=\text{dim}(G)$. Also $G'$ is smooth over $S$ of some relative dimension $g'$, so that the diagonal morphism, $$\Delta_{G'}:G'\to G'\times_S G',$$ is a regular embedding of codimension $g$. Thus, the base change graph morphism, $$\Gamma_f:G\to G\times_S G',$$ is also a regular embedding of codimension $g$.

Finally, for a flat morphism $X\to Y$ of relative dimension $d$ and for a regular embedding $Z\hookrightarrow X$ of codimension $e$, the flat locus of $Z\to Y$ contains every point of $Z$ where the fiber has dimension $d-e$. This follows from the local flatness criterion (use the Koszul complex of the regular sequence to resolve the ideal sheaf, and then use the codimension criterion for exactness of a Koszul complex to check exactness of the basechange of the Koszul complex to each fiber). Apply this with $Y=G'$, $X=G\times_S G'$ and $Z$ is the image of $\Gamma_f$.

  • 1
    $\begingroup$ Could we not also argue directly that $f$ is flat by the fiberwise criterion for flatness (since $G$ and $G'$ are flat locally of finite presentation over the base)? $\endgroup$ Sep 17, 2020 at 8:23

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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