If $A$ is a finite dimensional commutative, associative, unital algebra over a field $\mathbb{K}$ then does there exist a non-zero vector $v\in A\otimes_{\mathbb{K}}A$ such that $(a\otimes_{\mathbb{K}}1)v=(1\otimes_{\mathbb{K}}a)v$ for all $a\in A$?

If $A$ happens to be a Frobenius algebra then we can take $v$ to be the dual of the associated non-degenerate bilinear form. Moreover, I think I have shown that for a $\mathbb{K}$-basis of $A$ given by $e_{i}$ if there exists $v=\sum_{ij}\beta^{ij}e_{i}\otimes_{\mathbb{K}}e_{j}$ and $\beta^{ij}$ is non-degenerate then $\beta^{ij}$ gives a Frobenius form.

My main interest is the case where $\mathbb{K}=\mathbb{R}$ however I can't see this making much of a difference in general.

  • $\begingroup$ What about $v = 0$, or, more general, $v \in \mathbb{K}$? $\endgroup$
    – Dirk
    Jun 4, 2018 at 10:55
  • $\begingroup$ For $v\in\mathbb{K}-\{0\}$ we don't have $(a\otimes_{\mathbb{K}}1)v=(1\otimes_{\mathbb{K}}a)v$ for all $a\in A$ but I should have mentioned non-zero. $\endgroup$
    – Campbell
    Jun 4, 2018 at 11:14
  • $\begingroup$ @DirkLiebhold. For $v\in \mathbb{K}\setminus\{0\}$, typically $a\otimes_{\mathbb{K}}1\cdot v$ does not equal $1\otimes_{\mathbb{K}} a\cdot v$. Denote by $I$ the kernel of $A\otimes_K A\to A,$ $a\otimes b \mapsto a\cdot b$. Then $I$ is a nonzero ideal. By Nakayama, $I^\ell = I^{\ell+1}$ only if $I^\ell=0$. Since $A\otimes_K A$ is Artinian, some $I^\ell$ is zero, i.e., $I$ is nilpotent. Let $\ell$ be the largest integer such that $I^\ell$ is nonzero. Then for every nonzero $v\in I^\ell$, for every $a\otimes 1 - 1\otimes a$ in $I$, the product with $v$ equals $0$. $\endgroup$ Jun 4, 2018 at 11:20
  • $\begingroup$ Ah, yes, my error, sorry. $\endgroup$
    – Dirk
    Jun 4, 2018 at 11:54

1 Answer 1


Let $I$ be the ideal in $A\otimes A$ generated by elements of the form $a\otimes 1-1\otimes a$. (Equivalently, this is the kernel of the multiplication map $A\otimes A\to A$.) We will assume that $0\neq A\neq \mathbb{K}$, so that $0<I<A\otimes A$. As everything is finite-dimensional, the sequence of powers $I^n$ must eventually stabilise. Let $n$ be least such that $I^n=I^{n+1}=I.I^n$. By the standard determinant trick, this gives an element $u$ with $u=1\pmod{I}$ and $u.I^n=0$. Let $m$ be least such that $u.I^m=0$. As $u=1\pmod{I}$ we have $u\neq 0$, so $m>0$. Now $u.I^{m-1}\neq 0$ and $u.I^{m-1}\leq\text{ann}(I)$ so $\text{ann}(I)\neq 0$.

  • 1
    $\begingroup$ "the standard determinant trick" also known as Nakayama's lemma :) $\endgroup$ Jun 4, 2018 at 11:31
  • $\begingroup$ @darijgrinberg Matsumura calls that result Nakayama's Lemma, but many other people use that name just for the statement $\mathfrak{m}M=M\Rightarrow M=0$. I am no historian and do not really have an opinion. $\endgroup$ Jun 4, 2018 at 11:50
  • $\begingroup$ Suggestion: Nakayama's determinant trick :) Not a big deal, but I've gotten used to seeing at least one determinantal trick per proof, and those aren't all the same. $\endgroup$ Jun 4, 2018 at 11:57

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.