1

I would like to understand the tensor product of $A=\Bbb F_2(\sqrt{t})\otimes_{\Bbb F_2(t)}\Bbb F_2(\sqrt{t})$.

The extension $L/k:=\Bbb F_2(\sqrt{t})/\Bbb F_2(t)$ is a finite extension of degree $2$ purely inseparable since the minimal of $\sqrt{t}$ over $k$ is $P=X^2-t$ and any element $(a+b\sqrt{t})/(c+d\sqrt{t})\in L$ has its square in $k$. It must have nilpotents base on readings of different threads but if I compute the tensor product $L\otimes_kL$, with $L\cong k[X]/(X^2-t)$, I have:

$$A=L\otimes_kL=L\otimes k[X]/(X^2-t)=L[X]/(X^2-t)=L[X]/(X-\sqrt{t})^2$$

(edited: can't use CRT here!)

How could I find the nilpotents?

NevD
  • 525
  • Related: http://math.stackexchange.com/questions/345497/tensor-product-of-inseparable-fields. – user26857 Mar 11 '16 at 22:39
  • Indeed, you could have marked as duplicate but I didn't see this one. Anyway, I think Alex Wertheim's answer very usefull. I was tring some computations such as $P=a(\sqrt{t})+b(\sqrt{t})X$ nilpotent (!). Now it's clear. – NevD Mar 11 '16 at 22:49

1 Answers1

1

$A$ does have nilpotents, namely $X-\sqrt{t}$ by Frobenius (there is nothing mysterious here; this is simply what you have already observed). The equivalence you posted previously, namely $A = L[X]/(X-\sqrt{t})^{2} \cong L[X]/(X-t) \times L[X]/(X-t)$, was not correct; we cannot apply the Chinese remainder theorem here.

To compute nilradical $R$ of $A = L[X]/\langle (X-\sqrt{t})^{2} \rangle$, let $I = \langle (X-\sqrt{t})^{2} \rangle$ and recall that $R$ is the intersection of all prime ideals of $A$. The prime ideals of $A$ are in bijective correspondence with the prime ideals of $L[X]$ containing $I$ via the canonical quotient homomorphism $\pi \colon L[X] \to A$. It thus suffices to compute the intersection $J$ of all prime ideals of $L[X]$ containing $I$; then $\pi(J) = R$. Any prime ideal $\mathfrak{p}$ in $L[X]$ which contains $I$ must contain $X-\sqrt{t}$, hence $\langle X - \sqrt{t} \rangle$. On the other hand, $\langle X - \sqrt{t} \rangle$ is a prime ideal of $L[X]$ containing $I$, so we see $J = \langle X - \sqrt{t} \rangle$.

Alex Wertheim
  • 20,278
  • I suddenly realise that understand I can't use the chinese remainder! I edited. Could you detail the action of the Frobenius? – NevD Mar 11 '16 at 21:00
  • @happy: I have updated my answer with some computations which I believe answer your question. – Alex Wertheim Mar 11 '16 at 22:18