21

Let $(G,.)$ be a group and $m,n \in\mathbb Z$ such that $\gcd(m,n)=1$. Assume that $$ \forall a,b \in G, \,a^mb^m=b^ma^m,$$ $$\forall a,b \in G, \, a^nb^n=b^na^n.$$

Then how prove $G$ is an abelian group?

Some context: Some of these commutation relations often imply that $G$ is abelian, for example if $(ab)^i = a^i b^i$ for three consecutive integers $i$ then $G$ is abelian, or if $g^2 = e$ for all $g$ then $G$ is abelian. This looks like another example of this phenomenon, but the same techniques do not apply.

M.H
  • 11,498
  • 3
  • 30
  • 66

3 Answers3

19

Let $M \subset G$ be the subgroup generated by all $m$-th powers and let $N \subset G$ be the subgroup generated by all $n$-th powers. These subgroups are clearly abelian normal subgroups. Since $m$ and $n$ are coprime, $G = MN$, and hence $M \cap N$ is contained in the center $Z(G)$ of $G$. To prove that $G$ is abelian it suffices to show that $M$ and $N$ commute, that is $[M,N]=1$. Note that $[M,N] \subset M \cap N$ (since $M$ and $N$ are normal subgroups). Let $a \in M$ and $b \in N$. Then $[a, b] = a^{−1}b^{−1}ab \in M \cap N$. Hence $[a, b] = z$ with $z \in Z(G)$. Hence $b^{−1}ab = za$, whence $b^{−1}a^nb=z^na^n$. Since $a^n \in N$ it commutes with $b$, so $z^n=1$. Similarly $z^m=1$. Since $m$ and $n$ are relatively prime, we conclude $z=1$.

Nicky Hekster
  • 49,281
  • 1
    Dear Nicky! Let me ask sime questions: 1) Why $G=MN$? 2) Why $M\cap N$ is contained in the center $Z(G)$? – RFZ May 01 '18 at 14:21
  • 3
    (1) $gcd(m,n)=1$ whence you can find integers $k$ and $l$ with $1=km+ln$. So if $g \in G$ then $g=g^1=(g^k)^m \cdot (g^l)^n \in MN$. (2) Because $G=MN$ and $M$ and $N$ are abelian, any element from $M \cap N$ commutes with any element from $G$. Hope it is clear now. – Nicky Hekster May 01 '18 at 15:33
  • Dear Nicky! Thanks for answer! I got your solution and your approach is very simple and nice! +1 – RFZ May 01 '18 at 19:53
13

Write $mk+nl=1$ for some $k,l\in\mathbb Z$.

We have

$$\begin{align} ab&=a^{mk+nl}b^{mk+nl}\\ &=a^{mk}(a^{nl}b^{mk})b^{nl}\\ &\stackrel{(*)}=a^{mk}(b^{mk}a^{nl})b^{nl}\\ &=(a^{mk}b^{mk})(a^{nl}b^{nl})\\ &=(b^{mk}a^{mk})(b^{nl}a^{nl})\\ &=b^{mk}(a^{mk}b^{nl})a^{nl}\\ &\stackrel{(**)}=b^{mk}(b^{nl}a^{mk})a^{nl}\\ &=(b^{mk}b^{nl})(a^{mk}a^{nl})\\ &=b^{mk+nl}a^{mk+nl}\\ &=ba, \end{align}$$

where we used $$a^{nl}b^{mk}=b^{mk}a^{nl}\qquad(*)$$ and $$a^{mk}b^{nl}=b^{nl}a^{mk}\qquad(**)$$

Let's prove $(*)$ and $(**)$:

$$\begin{align} (a^mb^n)^{mk}&=a^m(b^na^m)^{mk-1}b^n\\ &=a^m(b^na^m)^{mk}a^{-m}, \end{align}$$

which implies

$$(a^mb^n)^{mk}a^m=a^m(b^na^m)^{mk}\\ \Rightarrow a^m(a^mb^n)^{mk}=a^m(b^na^m)^{mk},$$ hence $$(a^mb^n)^{mk}=(b^na^m)^{mk}.\qquad(\sharp)$$

Similarly we get $$(a^mb^n)^{nl}=(b^na^m)^{nl}.\qquad(\sharp\sharp)$$

Multiplying $(\sharp)$ and $(\sharp\sharp)$ we obtain $a^mb^n=b^na^m$.

Proceeding in the same way we get $a^nb^m=b^ma^n$

An easy induction shows now $(*)$ and $(**)$.

Shaun
  • 44,997
user26857
  • 52,094
-6

As $\gcd(m, n) = 1$, by Bezóut's identity there are $u$, $v$ such that $u m + v n = 1$.

Take $a, b \in G$ arbitrary. By the conditions, as $a^u \in G$ for all $u \in \mathbb{Z}$, and as the conditions allow to commute certain powers: $$ \begin{align*} a^{u m} b^{u m} &= b^{u m} a^{u m} \\ (a^{u m} b^{u m})^{v n} &= a^{u m + v n} b^{u m + v n} \\ &= a b \\ (b^{u m} a^{u m})^{v n} &= b^{u m + v n} a^{u m + v n} \\ &= b a \end{align*} $$ This because, e.g., $(a^{u m} b^{u m})^2 = a^{u m} b^{u m} a^{u m} b^{u m} = a^{u m} a^{u m} b^{u m} b^{u m} = a^{2 u m} b^{2 u m},$

since the middle pair commutes by the conditions. Now use induction.

(Perhaps the switching around can be expressed in a more compact way).

Nice problem!

amWhy
  • 209,954
vonbrand
  • 27,812
  • 1
    Great Solution :) thanx :) – FNH Mar 10 '13 at 19:52
  • 4
    @vonbrand:why we have $(b^{u m} a^{u m})^{v n} = b^{u m + v n} a^{u m + v n}$? – M.H Mar 10 '13 at 20:30
  • @MaisamHedyelloo , we have a in G , so $a^u$ is in G , choose u to be the same u in the relation $mu+vn=1$ , apply the hyposetes to $a_1$ = $a^u$ and use induction as it's mentioned in the answer in the speicial case for 2 . – FNH Mar 10 '13 at 20:45
  • 2
    How do you get $(a^{u m} b^{u m})^{v n} = a^{u m + v n} b^{u m + v n}$? Doesn't $(a^{u m} b^{u m})^{v n} = a^{u m v n} b^{u m v n}$? – Code-Guru Mar 10 '13 at 22:11
  • @MrWhy Huh? What's $a_1$? – Code-Guru Mar 10 '13 at 22:15
  • @Code-Guru , $a_1$ is a new element in the group , it's just a definition of a new element , $ (pq)^c $ = $(p^c q^c ) $ iff G is abelian " so we can't use this fact because we want to prove G is abelian ! . here , $( a^{um} b^{um} )^{vn}$ = $a^{um+vn} b^{um+vn}$ from the hyposthes in the question ,you can prove it easily by induction as mentined in the answer " and the answer contained a special case for 2 , which enable you to imagine the reason of this ! " – FNH Mar 11 '13 at 01:06
  • In the expression $(a^{um}b^{um})^{vn}$ there are $um\cdot vn$ copies of $a$ and the same number of copies of $b$. However in the claimed equal expression $a^{um+vn}b^{um+vn}$ there are only $um+vn$ copies each of $a$ and $b$. And certainly there is no integer identity saying that $um\cdot vn=um+vn$. The whole argument is thus based on misuse of properties of exponents. – coffeemath Mar 11 '13 at 05:04