5
$\begingroup$

Let $X$ be an affine algebraic set in $\mathbb C^n$, i.e., the vanishing set of a family of polynomials. Let's call a function $f:X\to \mathbb C$ locally rational if for every $x\in X$ there exists a Zariski open $U\subseteq \mathbb C^n$ such that the restriction of $f$ to $U\cap X$ is a rational function (i.e., given by a ratio of two polynomials).

If $X$ is irreducible, then there is a theorem in Shafarevich's book which says that any locally rational map $f:X\to \mathbb C$ is in fact a polynomial. Does anyone know of a counterexample to this statement for the case where $X$ is reducible? And in the latter case, what is the optimal refinement of the theorem?

$\endgroup$
2
  • 2
    $\begingroup$ If by "given by a ratio of two polynomials" you assume that the denominator does not vanish on $U\cap X$ (or, for that matter, on $U$), then the statement also holds for any affine algebraic set. This is built in scheme theory, under the fact that there is a sheaf on $\mathop{\rm Spec}(A)$ with value $A_f$ on $D(f)$. At the level of algebraic sets, unfortunately, Theorem 3.2 of Hartshorne's Algebraic Geometry assumes irreducibility. I had given a proof in general in my Alg. Geometry notes, Théorème 3.3.7, math.u-psud.fr/~chambert/publications/teach/Dea-1999.pdf $\endgroup$
    – ACL
    Mar 5, 2013 at 9:16
  • $\begingroup$ Thank you ACL! I am not an algebraic geometer, but for some reason I was not aware of this simple refinement of the standard argument. $\endgroup$
    – Valerie
    Mar 6, 2013 at 5:40

1 Answer 1

1
$\begingroup$

Answer: see the above instructive comment by ACL.

$\endgroup$
1
  • $\begingroup$ Dear @Valerie: You may consider accepting this answer to formally close the question. It may also be helpful to copy ACL's comment into the answer box, with proper attribution. $\endgroup$ Oct 23, 2013 at 16:43

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.