Conjecture. There exists a function $f:\mathbb{N} \rightarrow \mathbb{N}$ such that if $\alpha$ and $\beta$ are non-zero elements of the complex group algebra $\mathbb{C}[G]$ of a finite group $G$ such that $1\in \text{supp}(\alpha) \cap \text{supp}(\beta)$, $|\text{supp}(\alpha)|\leq |\text{supp}(\beta)|$ and $\alpha \cdot \beta =0$, then $\text{exp}\big (\langle \text{supp}(\alpha) \rangle \big) \leq f(|\text{supp}(\beta)|)$, where $\text{exp}(H)$ denotes the exponent of a finite group $H$ and $\text{supp}(\gamma)$ for an element $\gamma=\sum_{g\in G} \gamma_g \; g \in \mathbb{C}[G]$ denotes the set $\{ g\in G \;|\; \gamma_g \not=0\}$.

Motivation. If the conjecture is true, then the support of any zero divisor of the complex group algebra of any residually finite group generates a finite subgroup.

Proof. Let $\alpha$ be a non-zero element of $\mathbb{C}[G]$ for some residually finite group $G$ such that $\alpha \cdot \beta =0$ for some non-zero $\beta \in \mathbb{C}[G]$. Let $H=\langle \text{supp}(\alpha) \rangle$. By Zelmanov's celebrated result on restricted Burnside problem, it is enough to show that the exponent of $H$ is finite. Since $H$ is finitely generated residually finite, there exists a descending series $H=N_1\geq N_2 \geq \cdots$ of normal subgroups $H$ of finite index such that $\cap_{i\in\mathbb{N}} N_i=1$. Note that there exists $k\in\mathbb{N}$ such that $\bar{\alpha} \cdot \bar{\beta}=0$ in $\mathbb{C}[H/N_i]$ for all $i\geq k$ and $|\text{supp}(\bar{\beta})|=|\text{supp}(\beta)|=:t$, where $\bar{}$ is the natural ring epimorphism from $\mathbb{C}[H]$ onto $\mathbb{C}[H/N_i]$. Now if the above conjecture is true then $\text{exp}(H/N_i)\leq f(t)$ and so $\text{exp}(H)$ is finite. This completes the proof.

If the above proof and conjecture are true then the complex group algebras of torsion-free residually finite groups have no zero divisor.

  • $\begingroup$ There are finitely generated, infinite groups with finite exponent. So I think the argument is not quite complete, but for applications to torsion-free groups, this should not be a problem. $\endgroup$ Commented Oct 24, 2018 at 11:36
  • $\begingroup$ @SteffenKionke The condition ``residually finite" on a finitely generated group of finite exponent garantees that the group is finite. $\endgroup$ Commented Oct 24, 2018 at 15:30
  • $\begingroup$ Yes, thanks and sorry. I didn't pay attention... $\endgroup$ Commented Oct 24, 2018 at 16:10

1 Answer 1


The conjecture does not hold.

Let $m \geq 1$ be an arbitrary integer. Let $G = \langle x,y | x^m, y^5 , [x,y]\rangle$ be the finite abelian group isomorphic to a product of two cyclic groups $C_m \times C_5$. Let $\alpha = xy +y -x -1$ and let $\beta = y^4 +y^3 +y^2 +y +1$ in $\mathbb{Z}[G]$. Then $\alpha \beta = (x+1)(y-1)\beta= 0$. The support of $\beta$ has $5$ elements and the support of $\alpha$ has $4$ elements. Moreover, the support of $\alpha$ generates the group $G$, which has exponent $\mathrm{exp}(G) \geq m$.

  • $\begingroup$ Thanks! Actually I undertand from your example that I need a notion of ``irreducibilty" on a zero divisor something like the following: the element should not be written as the product of two group algebra element with support size less than the element and maybe this property that (one of them is a zero divisor). I will pose my question for the first unsettled case of existence of zero divisors with respect to the support size. $\endgroup$ Commented Oct 24, 2018 at 15:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.