4
$\begingroup$

I'm working on some palindromic words right now. Those are the elements of $F_2$, the free group on two generators, which are the same if reveresed. For example $xyx, y^2, xyxxyx$ and so on.

Can you write each element of $F'$, the derived group, as a product of palindromes of EVEN length? Does this property hold if you are not in $F$, but in a group $G$ in which you know that every element is a product of at most $n$ palindromes? If $G$ has this property then $G$ has finite palindromic width.

I've puzzled around quite a bit but can't find a way of even writing $x^{-1}y^{-1}xy$ as a product of palindromes of even length.

Thanks, Elisabeth

$\endgroup$
1
  • $\begingroup$ Since I initially got confused: "product of palindromes of even length" has to be understood as "product of (palindromes of even length)", and not as "product of an even number of palindromes", nor "product of palindromes, of even total length". $\endgroup$
    – YCor
    Commented Feb 16, 2023 at 10:11

2 Answers 2

14
$\begingroup$

The set of products of palindroms of even length $P$ is a normal subgroup, because the inverse of a palindrome is a palindrome and because $xpx^{-1} = xpx\cdot x^{-1} x^{-1}$ for all generators $x$ and all $p\in P$.

In the quotient $F(S)/P$ every generator $s\in S$ squares to 1. Therefore $F(S)/P$ is a quotient of $\langle S \mid \forall s\in S: s^2=1\rangle$. Since every palindrome of even length is of the form $x_1 x_2 \ldots x_{n-1} x_n x_n x_{n-1}\ldots x_2 x_1$, there are no further relations, so $F(S)/P \cong \langle S \mid \forall s\in S: s^2=1\rangle = \underbrace{C_2 \ast \ldots \ast C_2}_{|S|\,\text{times}}$. In particular the quotient is not abelian, if $|S|>1$. Hence $F(S)'$ is not contained in $P$.

$\endgroup$
9
$\begingroup$

Let $g=(1,2)$, $h=(2,3)$ two non-commuting involutions in (say) the symmetric group $S_3$. Consider the homomorphism $\varphi:F_2\to S_3$ with $\varphi(x)=g$, $\varphi(y)=h$.

If $w$ is any palindromic word of even length, then $w\in\ker(\varphi)$. However, $x^{-1}y^{-1}xy$ is not in $\ker(\varphi)$, and so can't be a product of palindromic words of even length.

$\endgroup$
3
  • $\begingroup$ Thanks! This also answers if it could be true if G was a soluble group! $\endgroup$ Commented Nov 19, 2013 at 18:03
  • $\begingroup$ This only works if the generators of $G$ don't have finite orders coprime to $2$. Could it still hold in some cases if we know that the generators of $G$ have finite orders coprime to $2$? $\endgroup$ Commented Nov 20, 2013 at 16:30
  • $\begingroup$ Sorry, yes of course then it holds because we can replace a palindrome of the form gxg by gx^{o(x)-1}g, where now o(x)-1 is even if o(x) is coprime to $2$, hence odd. $\endgroup$ Commented Nov 20, 2013 at 16:32

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .