What are the units in $R[X,X^{-1}]$, where $R$ is a commutative ring with $1$? I know that the question for polynomial rings is a standard textbook exercise. However, I couldn't find a reference for Laurent polynomials, since most people only seem to consider coefficients in an integral domain.

  • $\begingroup$ How does the standard proof for $R[X]$ go? We may be able to extend that proof, since $R[X,X^{-1}]$ is a localization of $R[X]$. $\endgroup$
    – Jose Brox
    Jul 8 '10 at 10:44
  • 13
    $\begingroup$ Look at Spec($R[x,1/x]) \rightarrow$ Spec($R$): fibers are Spec($k[x,1/x]$) with $k$ a field. So $\sum a_i x^i$ is a unit iff for all $p \in $ Spec($R$) some $a_i(p)$ is a unit & others vanish; i.e., some $a_i$ is a unit on Spec($R$) near $p$ and so others vanish near $p$. Thus, if $U_i$ is the open where $a_i$ is a unit then these are a separation of Spec($R$) and $a_j$ is nilpotent on $U_i$ for $j \ne i$. Answer: $R=\prod R_i$ with $a_i$ having unit $i$th component and nilpotent $j$th for $j \ne i$. So if Spec($R$) is connected and reduced then $R[x,1/x]=R^{\times} x^{\mathbb{Z}}$. $\endgroup$
    – Boyarsky
    Jul 8 '10 at 11:33
  • $\begingroup$ @MartinBrandenburg You mean $nil(R)$ instead of $rad(R)$. $\endgroup$ Dec 20 '15 at 11:58

You can find a more general result in the paper [1], which determines the units and nilpotents in arbitrary group rings $\rm R[G]$ where $\rm G$ is a unique-product group - which includes ordered groups. As the author remarks, his note was prompted by an earlier paper [2] which explicitly treats the Laurent case.

1 Erhard Neher. Invertible and Nilpotent Elements in the Group Algebra of a Unique Product Group
Acta Appl Math (2009) 108: 135-139

2 Ottmar Loos. Remarks on Holger P. Petersson's "Idempotent 2-by-2 matrices" http://homepage.uibk.ac.at/~c70202/jordan/archive/remarks/remarks.pdf


Thanks for your comments! A colleague just pointed me to an article by Karpilovsky ('On finite generation of unit groups of commutative group rings'). Translated into our setup, this gives the following result:

Take $r \in R^\times$, $a_i \in R$ nilpotent ($i \in \mathbb{Z}$), $k\ge 0$,$e_1,\dots,e_k \in R$ orthogonal idempotents that sum up to $1$, and fix $i_1,\dots,i_k \in \mathbb{Z}$. Then the element $$ r (1+\sum_i a_i X^i )(e_1X^{i_1} + \dots + e_kX^{i_k} ) $$ is a unit in $R[X,X^{-1}]$ and all units arise in this way.

In particular, all units are of the form $rX^j$ with $r \in R^\times$iff $R$ is reduced and connected, as Boyarsky pointed out.

I was hoping for a nice condition on the coefficients for the polynomial (as in the case $R[X]$). Maybe someone still sees how to simplify this statement or elegantly prove it in this setup?

  • 1
    $\begingroup$ That's exactly a concrete description of my comment/answer even in the general case (no connectedness or reducedness hypotheses): those idempotents correspond to a finite decomposition of $R$ into factor rings, and your unit in the $j$th factor ring of $R[X,1/X]$ is the product of the $j$th component of $r$ times $X^{i_j}$ times something which is 1 mod nilpotents. I think my proof in terms of fibering over Spec($R$) is elegant enough. :) $\endgroup$
    – Boyarsky
    Jul 8 '10 at 15:23
  • 2
    $\begingroup$ Um, thank you very much for the explanation! I have to admit I did not understand how to interpret your answer before. $\endgroup$
    – Seb
    Jul 8 '10 at 15:36
  • $\begingroup$ Thanks for the reference to that article, it may be useful for me! (btw, I think you should accept your own answer!) $\endgroup$
    – Jose Brox
    Jul 8 '10 at 22:48

A structure theorem and a sketch proof are presented in Exercise 3.7 of

The K-book: An introduction to algebraic K-theory (Graduate Studies in Mathematics 145) by Charles A. Weibel.

I found this reference very helpful. Indeed, the direct product decomposition of the unit group is made explicit there.


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.