4
$\begingroup$

Let $A$ be an abelian group with generators $x_1,x_2, \cdots, x_n$ and defining relations conssisting of $[x_i,x_j]$, $i<j=1,2, \cdots, n$, and $r$ further relations. If $r<n$, prove that $A$ is infinite.

Let $F$ be the free abelian group on $n$ generators. Do I have to prove the $r$ relations generate a finite subgroup of $F$? How to?

Thank you very much!

$\endgroup$
  • 5
    $\begingroup$ Consider the $r\times n$ matrix B, where $b_{ij}$ is the power of $x_j$ in the $i$th extra relation. Show there is a nonzero $n\times 1$ column vector of integers $v$ such that $Bv=0$. Perhaps use this to define a map from $A$ to $\mathbb{Z}$? $\endgroup$ – user641 Aug 9 '11 at 6:05
  • $\begingroup$ I don't understand the question. $\endgroup$ – Pierre-Yves Gaillard Aug 9 '11 at 6:33
  • $\begingroup$ @Pierre-Yves Gaillard: I just copied the problem from the book. I don't know how I can explain it clearer. Maybe the notation $[,]$ is not obvious? $[x_i,x_j]$ is the commutator of $x_i$ and $x_j$. $\endgroup$ – ShinyaSakai Aug 10 '11 at 13:34
  • $\begingroup$ @Steve D: Thank you very much. Please allow me to write it more detailed. I hope I am not wrong. ||||| Write $A$ in the addititive form. The $n$ generators are $x_j,j=1, \cdots, n$, and the $r$ relators are $y_i=\sum_{j=1}^nb_{ij}x_j,i=1, \cdots, r$, $b_{ij} \in \mathbb{Z}$. Let the matrix $B=(b_{ij})_{r \times n}$. As $r<n$, the rank of $B$ is less than $n$. There exists a nozero column vector $v=(v_1, \cdots, v_n)^T$ with all the $v_i$'s in $\mathbb{Z}$, such that $Bv=0$. $\endgroup$ – ShinyaSakai Aug 10 '11 at 13:36
  • $\begingroup$ Let $\phi: A \rightarrow \mathbb{Z}$, $\sum_{j=1}^nk_jx_j \mapsto \sum_{j=1}^nk_jv_j$. It is clear that $\phi$ maps the relators to $0$, and is a morphism between the two additive groups. As $v \neq 0$, there is some $v_l \neq 0, 1\leq l \leq n$, then $\phi(A) \supseteq v_l\mathbb{Z}$, so $|A| \geq |v_l\mathbb{Z}|$. $A$ is an infinite group. $\endgroup$ – ShinyaSakai Aug 10 '11 at 13:36
2
$\begingroup$

You can use linear algebra to explicitly find an element of $A$ with infinite order:

Suppose $A$ is generated by $x_1,\ldots,x_n$ subject to relations $$ m_{i1}x_1 + \cdots m_{in}x_n = 0,\quad i=1,\ldots,r<n. $$ Let $M=(m_{ij})\in R^{r\times n}$.

Let $F$ be the free abelian group on the $n$ generators $x_1,\ldots,x_n$. Then there is a unique surjection $\phi:F\to Z^n/C_Z(M^T)$ such that $\varphi(x_i)=e_i$ for all $i=1,\ldots,n)$, where $e_i$ is the $i$-th standard basis vector of $Z^n$ and $C_Z(M^T)$ denotes the integral column space of $M^T$, i.e., the set of all $Z$-linear combinations of columns of $M^T$. By design, $\phi$ vanishes on the subgroup $G$ of $F$ generated by the set $$\{m_{i1}x_1+\cdots + m_{in}x_n:i=1\ldots,r\}.$$ Therefore, $\varphi$ descends to a homomorphism $$\bar{\varphi}:A=F/G\to Z^n/C_Z(M^T).$$ Since $\varphi$ is surjective, so is $\bar{\varphi}$. Thus, it suffices to show that $Z^n/C_Z(M^T)$ is infinite.

Since $M$ has fewer rows than columns (here we use the inequality $r<n$), $$ \operatorname{dim}N(M)> 0\quad(\text{dimension over }\mathbb{Q}). $$ Here, $N(M)$ means nullspace over $\mathbb{Q}$. Let $x\in N(M)$ be nonzero and let $k$ be a integer such that $y :=kx\in Z^n$. Therefore, $Zy$ is an infinite subset of $N(M)\cap Z^n$. By a fundamental theorem of linear algebra, $N(M)$ is the orthogonal complement of $C(M^T)$ (column space over $\mathbb{Q}$). In particular, $N(M)\cap C(M^T)=\{0\}$. It follows that $$Zy \cap C_Z(M^T) \subseteq Zy\cap C(M^T) = \{0\}.$$ This implies that the restriction $$\phi|_{Zy}:Zy\to Z^n/C_Z(M^T)$$ is injective. Since $Zy$ is infinite, $Z^n/C_Z(M^T)$ must be infinite, too, as was to be shown.

Remark: By taking a basis of $N(M)$ (row reduction!) and clearing denominators, rather than just a single nonzero vector, you can conclude that $A$ has rank at least $n-\operatorname{rank}(A)$ by producing an explicit subgroup of that rank.

| cite | improve this answer | |
$\endgroup$

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.