Let $G,H$ be torsion-free abelian groups such that $k[G] \cong k[H]$ for any field $k$. Then is it true that $G \cong H$ ?

If this is not true, then what if I change the hypothesis to $R[G]\cong R[H]$ for any non-zero commutative unital ring; is the conclusion true then ?

  • $\begingroup$ I wonder what happens if one asks the same question but $k$ is fixed. $\endgroup$ Jun 27 '17 at 17:47
  • $\begingroup$ @Max : true , I wonder that too . $\endgroup$
    – user
    Jun 27 '17 at 18:06
  • $\begingroup$ @Max : by Eric's answer , $k[G]^\times \cong k^\times \times G$ , so for a fixed $k$ , we atleast have $k^\times\times G \cong k^\times \times H$ ; where $k^\times , G,H$ are all abelian groups , so we can infer that $G \cong H$ whenever $k^\times $ has the cancellation property in the category of abelian groups ; in particular when $k^\times$ is finitely generated , so at least when $k$ is finite $\endgroup$
    – user
    Jun 27 '17 at 18:12
  • $\begingroup$ @Max : so in particular , having $k[G]^\times \cong k^\times \times G$ is a desirable property for our conclusion to hold ; and it is a conjecture of Kaplansky (the Units Conjecture) that whenever $k$ is a field and $G$ is torsion free , then $k[G]^\times \cong k^\times \times G$ $\endgroup$
    – user
    Jun 27 '17 at 18:24
  • $\begingroup$ But this won't be enough : there is no cancellability property for arbitrary groups (although there is one for finite groups, but here $G$ is torsion free) $\endgroup$ Jun 27 '17 at 18:35

The group of units of $k[G]$ is exactly $k^\times\times G$. In particular, if $k=\mathbb{F}_2$, then the group of units is $G$, so $G$ can be recovered from the ring $\mathbb{F}_2[G]$.

To prove this, note that any torsion-free abelian group $G$ admits a total ordering compatible with the group structure (choose a totally ordered basis for $G\otimes\mathbb{Q}$ and use the lexicographic order). Considering $k[G]$ to be $G$-graded, suppose $u\in k[G]$ is a unit with inverse $v$. Let $a$ be the minimum degree of a homogeneous part of $u$ and $b$ be the maximum degree, and let $c$ be the minimum degree of a homogeneous part of $v$ and $d$ be the maximum degree. Then the product $uv$ has homogeneous parts of degree $a+c$ and also $b+d$. Since $uv=1$, we must have $a+c=b+d$. Since $a\leq b$ and $c\leq d$, this is only possible if $a=b$ and $c=d$, which means $u$ and $v$ are homogeneous. That is, $u$ is a unit of $k$ times an element of $G$.

  • $\begingroup$ what is the grading that you give to $k[G]$ ? $\endgroup$
    – user
    Jun 27 '17 at 17:15
  • $\begingroup$ A homogeneous element of degree $g$ is just anything of the form $ag$ for $a\in k$. $\endgroup$ Jun 27 '17 at 17:52
  • $\begingroup$ also I have made some comments following this question of mine , in reply to a comment of Max . Could you please enlighten more ? Thanks again $\endgroup$
    – user
    Jun 27 '17 at 18:41
  • $\begingroup$ Related to this MO question : do you know what happens if we assume that $k(G) \cong k(H)$ for any field $k$? (Here $k(G)$ denotes the field of fractions of $k[G]$, whenever $G$ is a torsion-free abelian group). $\endgroup$
    – Watson
    Jun 27 '17 at 22:02

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.