Let $R$ be a right (and left) Noetherian ring and $T=R[x]$ its polynomial ring. It was shown by Stafford that the set $S=\{f\in T\:|\:f\text{ monic}\}$ is a right denominator set. So my question is, if $g\in S$, then is $G=\{g^a\:|\:a\geq 0\}$ a right denominator set also?

If $t\in T$ and $g^i$ then I want to show $tG\cap g^i T\neq\emptyset$. As $g^i\in S$ then I know there exists some $t^\prime\in T$ and $f\in S$ such that $g^i t^\prime=tf$. So the question then becomes, does $f=g^j$ for some $j\geq 0$?

Edit: I thought of using the following result:

If $A\subset B$ is a multiplicative subset of a ring $B$ such that $aB$ is a two-sided ideal for any $a\in A$, then $A$ is a right permutable set.

Now, for $g^i\in G$ it is clear that $g^iT$ is right ideal of $T$. So, let $t\in T$ then $tg^iS\cap g^iT\neq\emptyset$, i.e. for some $s\in S$ and $t^\prime\in T$ we have $tg^is=g^it^\prime$. This seems promising but then I am unsure of how to proceed. I need to show $tg^i\bar{t}\in g^iT$ but do we need $\bar{t}=s\bar{t}^\prime$?


1 Answer 1


No, this is not true in general.

Let $g = x+\alpha$, let $t= \beta$ be an invertible element . Then we need to find $t'$ in $R[x]$ such that $(x+\alpha) t' = \beta (x+\alpha)^j = (x+ \beta \alpha \beta^{-1})^j \beta $, so $(x+\alpha) (t' \beta^{-1}) =(x+\beta \alpha \beta^{-1})^j$.

In particular, take $R$ to be a matrix algebra with $\alpha = \begin{pmatrix} 1 & 0 \\ 0 & 0 \end{pmatrix}$ and $\beta = \begin{pmatrix} 0 & 1 \\ 1 & 0 \end{pmatrix}$, so $\beta \alpha \beta^{-1}$ is $\begin{pmatrix} 0 & 0 \\ 0 & 1 \end{pmatrix}$.

If $t' \beta^{-1}$ satisfies this identity, then it must be a diagonal matrix, as any non-diagonal matrix polynomial will have nonzero off-diagonal entries when multiplied by $x+\alpha$ (by looking only at the highest-degree non-diagonal entries and multiplying them by $x$ to get the highest-degree non-diagonal entries of the product, we can see that there is no cancellation in the top degree).

So $t'$ if it exists is a diagonal matrix, which implies that $x+0$ divides $(x+1)^j$ in the ring of polynomials in one variable, which is false, and thus $t'$ does not exist.

  • $\begingroup$ Thanks for the answer, Will. You mention group algebras. What if $R$ is a group algebra over a finite group? $\endgroup$ Aug 3, 2017 at 22:08
  • $\begingroup$ @SamWilliams: I can make this counter example work, by a similar argument, in a 2 x 2 matrix algebra, which of course occurs in finite group algebras. $\endgroup$
    – Will Sawin
    Aug 4, 2017 at 1:40
  • $\begingroup$ @SamWilliams The counterexample is now simplified and applies to the group algebras of, e.g., $D_4$. $\endgroup$
    – Will Sawin
    Aug 5, 2017 at 0:37

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.