MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

We say that a group $(A, \cdot)$ is bi-orderable if there exists a total order $\preceq$ on $A$ such that $xz \prec yz$ and $zx \prec zy$ for all $x,y,z \in A$ with $x \prec y$.

Let $m,n$ be non-zero integers, and let ${\rm BS}(m,n)$ denote the Baumslag-Solitar group $\langle a, b \mid a^{-1} b^m a = b^n\rangle$. I was happily surprised to learn from Yves Cornulier (see here) that ${\rm BS}(1,n)$ is bi-orderable for $n \ge 1$, which naturally leads to the following:

Q1. Is there a ``nice'' characterization of those pairs $(m,n)$ such that ${\rm BS}(m,n)$ is bi-orderable?

E.g., ${\rm BS}(m,n)$ is not bi-orderable if $mn < 0$ (wlog, say that $1 \prec b$. Then, $b^{-1} \prec 1$, with the result that $ 1 \prec a^{-1} b^m a$ and $b^n \prec 1$ for $m > 0$, and dually $a^{-1} b^m a \prec 1$ and $1 \prec b^n$ otherwise) or $m = n \ge 2$ (wlog, say that $ab \prec ba$. Then, $a^n b \prec a^{n-1} b a \prec \cdots \prec aba^{n-1} \prec ba^n$). But what about the other cases?

Update on Q1. The question has been completely answered: ${\rm BS}(m,n)$ is bi-orderable if and only if $mn > 0$ and $\min(|m|,|n|) = 1$. For details, see Yves' answer below.

On a related note, let ${\rm D}(m,n)$ be the two-generator one-relator group $\langle a, b\mid a^{-1} b a^m = b^n\rangle$. This is sort of a variant of ${\rm BS}(m,n)$, and, while different in many ways, I'd like to know if both have a similar behavior as for orderability (by the way, does ${\rm D}(m,n)$ have a conventional name, so that I can look for references by myself?). In particular:

Q2. Is there a ``nice'' characterization of those pairs $(m,n)$ such that ${\rm D}(m,n)$ is bi-orderable?

In fact, ${\rm BS}(1,n) = {\rm D}(1,n)$, so the above (and, especially, Yves' argument) proves that ${\rm D}(1,n)$ is bi-orderable if and only if $n \ge 1$. But what about the other cases? E.g., is ${\rm D}(n,n)$ bi-orderable for $n \ge 2$?

Here is a question similar to Q1 (with total orders replaced by partial ones).

share|cite|improve this question
If $m,n\ge 2$ are not coprime then $BS(m,n)$ is not bi-orderable. Indeed it contains the amalgam $A(m,n)=\langle x,y\mid x^m=y^n\rangle$ as a subgroup. Then $A(m,n)$ is not bi-orderable. Indeed write $m=sp,n=sq$ with $s,p,q\ge 2$, $X=x^p$ and $Y=x^q$. Then $X\neq Y$ and $X^s=Y^s=1$. But extraction of $s$-roots is unique in a bi-orderable group. Indeed if $X<Y$, then $YX^{-1}>1$, hence $1<Y^{s-1}(YX^{-1})Y^{-s+1}=Y^sX^{-1}Y^{-s+1}=X^{s-1}Y^{-s+1}$, hence $X^{s-1}>Y^{s-1}$, but also since $X<Y$ we have $X^{s-1}<Y^{s-1}$, contradiction, and similarly contradiction if $Y>X$. – YCor Oct 31 '13 at 12:45
Also $BS(2,n)$ is not bi-orderable for any $n\ge 2$, because $A(2,n)$ is not bi-orderable. The reason is that in a biorderable group, $[X^2,Y]=1$ implies $[X,Y]=1$. Indeed $X(XYX^{-1}Y^{-1})X^{-1}=YXY^{-1}X^{-1}=(XYX^{-1}Y^{-1})^{-1}$, i.e. $c=XYX^{-1}Y^{-1}$ is conjugate to is inverse, and hence is trivial. This maybe generalizes to showing $[X^m,Y]=1\Rightarrow [X,Y]=1$ for any $m\ge 2$, in which case it would follow that $A(m,n)$ and hence $BS(m,n)$ is never bi-orderable if $m,n\ge 2$. – YCor Oct 31 '13 at 13:17
Many thanks for the interest, Yves. Just a minor detail as for your 1st comment: I think it should be $Y = y^q$, so $X^s = Y^s = 1$ isn't true. Nevertheless, $X^s=Y^s$ is enough to conclude: Since $X\ne Y$ (here is where we use that $s =\gcd(m,n) \ge 2$), then either $X\prec Y$, and then $X^s\prec Y^s$, or $Y \prec X$, and then $Y^s\prec X^s$. – Salvo Tringali Oct 31 '13 at 13:29
As for your 2nd comment: $[X^m, Y] = 1$ iff $X^m Y = Y X^m$, and this is impossible if $m \ge 2$ for the same reason expressed in the OP with reference to the orderability of ${\rm BS}(n,n)$ when $n \ge 2$. – Salvo Tringali Oct 31 '13 at 13:34
If $XY \ne YX$, then $X^mY=YX^m$ is impossible for $m \ge 1$. For suppose wlog $XY \prec YX$. Then, $X^m Y \prec X^{m-1} Y X \prec \cdots \prec XYX^{m-1} \prec YX^m$, which is the same argument used in the OP to prove that ${\rm BS}(n,n)$ is not bi-orderable for $n \ge 2$ (it works as well for $|n| \ge 2$). – Salvo Tringali Oct 31 '13 at 13:44
up vote 9 down vote accepted

(After the discussion in the comments.) $BS(m,n)$ is bi-orderable iff $mn>0$ and $\min(|m|,|n|)=1$.

Since it was already mentioned that $mn<0$ implies $BS(m,n)$ not bi-orderable and that $BS(1,n)$ is bi-orderable for all $n\ge 2$, all remains is to check that $2\le m\le n$ implies $BS(m,n)=\langle t,x\mid tx^mt^{-1}=x^n\rangle$ is not bi-orderable.

Define $y=txt^{-1}$. Then $y^m=x^n$. So $x$ commutes with $y^m$. So (*) $x$ commutes with $y$. But this is not the case.

(*) in a biorderable group, as mentioned by Salvo, $[X,Y^m]=1$, $m\ge 1$ implies $[X,Y]=1$, because otherwise if $XY<YX$, $X^mY<X^{m-1}YX<\dots<YX^m$ and similarly $YX<XY$ implies $YX^m<X^mY$.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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