1
$\begingroup$

Let $R$ be a commutative ring whose characteristic is either prime or $0$, such that $R/N$ is an integral domain, where $N$ is the nilradical, and $p: R \rightarrow R/N$ the canonical map. Is there a ring map $g: R/N \rightarrow R$ such that $p \circ g$ is the identity on $R/N$?

In geometric terms (since I really only care about finitely generated $k$-algebras for this): let $X$ be a(n affine) irreducible $k$-scheme, $Y=X_{\text{red}}$ its reduced structure (which is then a $k$-variety), and $j: Y \rightarrow X$ the natural inclusion. Is there a regular map $f: X \rightarrow Y$ such that $f \circ j$ is the identity on $Y$?

The reason for the assumption on characteristic is to avoid silly examples like $\mathbb Z/8\mathbb Z$.

$\endgroup$
3
$\begingroup$

No, that is not true. Consider the commutative ring $$R=k[x,y,z,t]/\langle xy+z^2-t,t^2\rangle.$$ The nilradical is $$N= \langle xy+z^2-t,t \rangle/\langle xy+z^2-t,t^2 \rangle.$$ Of course the quotient is $$R/N = k[x,y,z,t]/\langle xy+z^2,t \rangle = k[x,y,z]/\langle xy+z^2 \rangle.$$ Denote by $$q:k[x,y,z]\to k[x,y,z]/\langle xy+z^2 \rangle,$$ the standard quotient homomorphism. As an $R/N$-module, the nilradical is $$ N = \overline{t} k[x,y,z]/\langle xy+z^2 \rangle.$$

Now, for a given $k$-algebra homomorphism, $$ \widetilde{g}:k[x,y,z]\to R, $$ if $p\circ \widetilde{g}$ is the quotient $q$, then $$ \widetilde{g}(x) = \overline{x} + a\overline{t}, \ \widetilde{g}(y) = \overline{y} + b\overline{t}, \ \widetilde{g}(z) = \overline{z} + c\overline{t}. $$ Since $xy+z^2$ is in the kernel of $q$, in order for $\widetilde{g}$ to factor through $q$, we must have the following relation in $N$, $$ 0=\widetilde{g}(xy+z^2) = (\overline{x}+a\overline{t})(\overline{y}+b\overline{t}) + (\overline{z}+c\overline{t})^2 = \overline{t} + (b\overline{x} + a\overline{y} + 2c\overline{z})\overline{t}.$$ Since $b\overline{x}+a\overline{y}+2c\overline{z}$ is in the maximal ideal $\langle x,y,z \rangle/\langle xy+z^2 \rangle$ and $1$ is not in that maximal ideal, there is no choice of $a$, $b$ and $c$ so that both $p\circ \widetilde{g}$ equals $q$ and $\widetilde{g} = g\circ q$ for some choice of $g$.

An excellent resource for this sort of thing is Mike Artin's "Lectures on Deformations of Singularities".

$\endgroup$

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.