2
$\begingroup$

For any embedding of smooth varieties $X\subset Y$ given by an ideal sheaf $I\subset \mathcal {O}_Y$, it is well known that the normal cone $C _{X/Y}= \textbf{Spec}(\oplus_{\geq 0} I^i/I^{i+1})$ is isomorphic to the normal bundle of $X$ to $Y$. Therefore, its fiber over each point $x \in X$ is the affine space associated to the quotient $T_xY/T_xX$ of the tangent space of $Y$ by the tangent space of $X$ at $x$.

Now let $X$ be an arbitrary subscheme of a smooth variety $Y$. Suppose further that over any point $x\in X$ the fiber $(C _{X/Y})_x$ of the canonical morphism $C _{X/Y}\to X$ over $x$ is isomorphic to the affine space $T_xY/T_xX$.

Question: Is it true that X must be smooth?

For example, if, in addition to the above assumptions, both $X$ and $C _{X/Y}$ are irreducible varieties, then the answer is positive. Indeed, \begin{gather}dim (C _{X/Y})_x\geq dim C _{X/Y} -dim X \\= dim Y - dim X\geq dim T_xY -dim T_xX \end{gather} so all inequalities must be equalities, hence $dim T_xX =dim X$ for all $x$, i.e. $X$ is smooth.

$\endgroup$
0
$\begingroup$

Yes, $X$ must be smooth. The question is local at $x \in X$. Write $R$ for the local ring $\mathcal{O}_{Y,x}$, and $\mathfrak{m}$ for the maximal ideal of $R$. Write $I$ for the ideal defining $X$. Then $(C_{X/Y})_x = \mathrm{Spec}(\oplus I^n/mI^n)$. Its Krull dimension is called the analytic spread of $I$; it is at least $\mathrm{height}(I)$. (See the chapter on "Minimal reductions" in the book by Swanson and Huneke on integral closure.) On the other hand, $\dim T_xX = \dim_{R/\frak{m}}(\frak{m}/(I + \frak{m}^2)) \geq \dim R/I$. Since $\dim R = \dim R/I + \mathrm{height}(I)$, the given hypothesis implies that $\dim T_xX = \dim R/I$, i.e., $R/I$ is a regular local ring.

$\endgroup$
  • $\begingroup$ Thank you very much. $\endgroup$ – sky223 Oct 15 '19 at 21:54

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.