Let $A$ and $B$ be noetherian normal rings and let $f:A\rightarrow B$ be a finite but non-flat ring homomorphism. We can also assume $Spec(A)$ connected if necessary. We put on $B$ the structure of $A$-module given $f$.

Is it true that $B\otimes_A B$ is normal?

The example I need to tackle comes from the following situation: take $A$ normal and noetherian, $K$ its function field, $K\subset L$ a finite extension, $B$ the normalization of $A$ in $L$.


This is false, even with flatness.

Say $k = \mathbb{F}_p$ for a prime number $p$, $A = k[t]$, $B = k[s]$, and $f(t) = s^p$. Then $A$ and $B$ are noetherian, normal, connected, and $B = A[x]/(x^p - t)$, which is finite flat over $A$. However, $B \otimes_A B = A[x,y]/(x^p - t, y^p - t)$, which carries is nilpotent element $x-y$ as $(x-y)^p = x^p - y^p = 0$.

[The same example, in characteristic $0$, shows that $B \otimes_A B$ need not be a domain.]

  • $\begingroup$ I am confused. Assume $k$ algebraically closed. Why does your example not contradict EGA IV, 6.14.1 where we take $Y=Spec(A)$, $Y^{\prime}=Spec(B)=X$? $\endgroup$
    – jikj
    Jan 19 '14 at 20:30
  • $\begingroup$ The map $f$ in the answer does not satisfy the hypotheses in EGA (look at the $t=0$ fibre). $\endgroup$
    – anonymous
    Jan 19 '14 at 21:14

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.