1
$\begingroup$

Let $S$ an arbitrary scheme and denote by $\Delta: S \to S \times S$ the diagonal immersion and $p_i: S \times S \to S$ the both projections to first resp second factor. (in following we will wlog talk only the projection on first factor, ie we set $p:=p_1$)

In general when $S$ isn't separated the image $\Delta(S) \subset S$ is only locally closed and not closed in $S \times S$. Since the restricted projection $p \vert _{\Delta(S)}: \Delta(S) \to S$ is an isomorphism that's not interesting to study this map. But the extended restriction

$$p \vert _{\overline{\Delta(U)}}:\overline{\Delta(U)} \subset U \times U \to U$$ looks more interessant. Which scheme theoretic map properties $\mathcal{Q}$ does $p \vert _{\overline{\Delta(U)}}$ inherit from $p \vert _{\Delta(S)}$? (ie $\mathcal{Q}$ stands for example beeing etale, smooth, flat etc)

Indeed the main motivation behind this question is the question if $p \vert _{\overline{\Delta(U)}}$ is always etale? (see This closely related question & discussion). Alexl gave the example where $S$ is the affine line with doubled origin and in this case it's quite obvious that $p \vert _{\overline{\Delta(U)}}$ is etale as well. But does this example cover all phenomena can occure by passing from $\Delta(S)$ to $\overline{\Delta(S)}$ as subscheme of $S \times S$?

More generally assume that $f: X \to Y$ a morphism of schemes and $S$ is locally closed subscheme of $X$ and $\overline{S}$ it's schematic closure in $X$. Assume the restriction $p \vert _S:S \to Y$ is etale (or has another fancy morphism property $\mathcal{Q}$ but I'm primary interested on etaleness as archetypical property), does the "extended" restriction $p \vert _{\overline{S}}:\overline{S} \to Y$ has also property $\mathcal{Q}$?

Is there any criterion / theory treating this problem and probably gives an answer for which morphism properties $\mathcal{Q}$ the above behaves well?

$\endgroup$
3
  • 1
    $\begingroup$ Take a degree $d \geq 3$ covering of smooth projective curves $f \colon X \to Y$, branched at a single point $y \in Y$ and such that $$f^{*}(y)=kx_0 + x_1 + \ldots + x_{d-k},$$ with $k <d$. Then the restriction $$f \colon X-\{x_0 \} \to Y$$ is surjective and unramified, hence étale, but it extension to $X$ clearly is not. $\endgroup$ Commented Jul 16, 2020 at 14:19
  • $\begingroup$ So the attempt to generalize the issue was too optimistic. But in more concrete situation when $f$ is the diagonal map $\Delta: S \to S \times S$. Why is the restriction $p_1 \vert _{\overline{\Delta(S)}}:\overline{\Delta(S)} \subset S \times S \to S$ of the projection $p_1: S \times S \to S$ nevertheless etale? The issue that I wanted to understand here mathoverflow.net/questions/365735/… that a usual scheme $S$ is an algebraic space. So we need data $(U,R)$ with $U:=S$ and $R \subset U \times U$ closed (!). $\endgroup$
    – user267839
    Commented Aug 10, 2020 at 14:48
  • $\begingroup$ The natural choice for $R$ is $R:=\overline{\Delta(U)}$ since if $S$ is not separated, the image $\Delta(U)$ isn't closed. That is we need etaleness of projection maps $p_i: \overline{\Delta(U)} \to U$. Do you know an argument why that's true? $\endgroup$
    – user267839
    Commented Aug 10, 2020 at 14:48

0

You must log in to answer this question.

Browse other questions tagged .