Let $K/k$ be an arbitrary field extension and $X$, $Y$ varieties over $k$ (lets assume projective and perhaps smooth to avoid technicalities). There is a fine moduli space of morphisms between $X$ and $Y$ parametrized by a scheme $Hom(X,Y)$ over $k$. My question is whether $Hom(X,Y)_K$ is isomorphic to $Hom(X_K, Y_K)$ where ${}_K$ denotes tensoring with $Spec K$. From the universal property of $Hom(X,Y)$ we have a bijection between maps $Spec K\to Hom(X,Y)$ (i.e. $K$-rational points, as a set contained in $Hom(X,Y)_K$) and maps $Y_K\to X_K$. I seem to be a bit confused as to whether this is enough to conclude.

Finally, could we get a similar thing to work for the coarse Kontsevich moduli space $\mathcal{M}_{g,n}(X,\beta)$ even though there is no universal family? Thanks.

  • 1
    $\begingroup$ It is not correct that the Hom scheme is projective. It need be neither quasi-compact nor universally closed. Rather it is a countable union of quasi-projective schemes. One example is where both $X$ and $Y$ equal the projective line $\mathbb{P}^1$. Yes, $\textit{Hom}_k(X,Y)_K$ equals $\textit{Hom}_K(X_K,Y_K)$. No, the same is not true for the Kontsevich moduli space. $\endgroup$ – Jason Starr Jun 30 '12 at 14:04
  • $\begingroup$ Apologies, I did not even intend to have the word projective in there, corrected! $\endgroup$ – Nadal Jun 30 '12 at 14:55
  • $\begingroup$ tea.mathoverflow.net/discussion/1392/… $\endgroup$ – Martin Brandenburg Jun 30 '12 at 15:01

If $X,Y$ are $S$-schemes, then $\underline{\hom}_S(X,Y)$ denotes the sheaf on $S$-schemes defined by $T \mapsto \hom_T(X_T,Y_T)$. Now, if $S'/S$ is any base change, we have for every $S'$-scheme $T$:

$$\hom_{S'}(T,\underline{\hom}_S(X,Y)_{S'}) = \hom_S(T,\underline{\hom}_S(X,Y)) = \hom_T(X_T,Y_T)$$

$$ = \hom_T((X_{S'})_T,(Y_{S'})_T)) = \hom_{S'}(T,\underline{\hom}_{S'}(X_{S'},Y_{S'}))$$

Hence, $\underline{\hom}_S(X,Y)_{S'} = \underline{\hom}_{S'}(X_{S'},Y_{S'})$.

  • 3
    $\begingroup$ Here is a counterexample for the Kontsevich space. Start with a smooth, plane conic $C$ in $\mathbb{P}^2_k$ such that $C$ has no $k$-rational point, e.g., the conic with defining equation $X_0^2 + X_1^2 + X_2^2$ over $\mathbb{R}$. Let $D$ be an effective, degree $2$, Cartier divisor in $C$, e.g., the common locus of $X_0$ and $X_1^2 + X_2^2$ for the above conic. There is a $k$-point of $\mathcal{M}_{0,0}(\mathbb{P}^2,4)$ "parameterizing" a double cover $f:B\to C$ branched over $D$. However, $f_*\mathcal{O}_B/\mathcal{O}_C$ has degree 1, so $f$ is not defined over $k$. $\endgroup$ – Jason Starr Jun 30 '12 at 16:54
  • $\begingroup$ Typo correction: $f_*\mathcal{O}_B/\mathcal{O}_C$ has degree $-1$, not degree $+1$. The dual has degree $1$ -- this is impossible since $C$ has no $k$-rational point, thus $f$ is not defined over $k$. $\endgroup$ – Jason Starr Jun 30 '12 at 17:00
  • $\begingroup$ Jason, this is not a comment to my trivial, but rather a profound answer by your own. Post it as an answer, please. tea.mathoverflow.net/discussion/1392/… $\endgroup$ – Martin Brandenburg Jun 30 '12 at 18:04

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.