2

Here my question, does there always exist such groups in which product of two elements (namely , $a,b$ with $|a|=m,|b|=n$) with order $k$? Then what will be the structure of that group?

Here $m,n,k$ may distinct or not but runs over all natural numbers.

I can think for just of some particular cases, can't think arbitrarily.

If $m=2,n=2$ then for every $k$ we will have dihedral group $D_k$(with order $2k$) in our hand.

If $m=2,k=2$ then for every $n$ we will have dihedral group $D_n$ in our hand.

.........

Is there any suitable way to think?

Do those question have an affirmative answer?

Shaun
  • 47,747
A learner
  • 2,871
  • Well, define $G:=\langle a,b|a^m=b^n=(ab)^k=e\rangle$ and find out the order of $a$, $b$ and $ab$. – Mastrem Jun 17 '21 at 15:00
  • 1
    @Mastrem In your given example, how could you know order of $a,b,ab$ be $m,n,k$ respectively always? – A learner Jun 17 '21 at 15:08
  • Is it not that in general some case could imply $a,b$ to trivial? – A learner Jun 17 '21 at 15:10
  • 2
    Please read the Wikipedia article Presentation of a group for the notation $G:=\langle a,b|...\rangle$. – Somos Jun 17 '21 at 15:11
  • 4
    @Mastrem This is not the way to approach the question, because there is no obvious way prove that the elements of the group you define have the "correct" orders. Essentially, you have rephrased the problem into one which is no easier. – user1729 Jun 17 '21 at 15:19
  • Note that this has nothing to do with [tag:order-theory]. – Shaun Jun 17 '21 at 15:20
  • @Somos Ok, for example ,when we took such generated group example like ${\langle(a,b)\rangle: a^{4}=1=b^{3},ab=b^{2}a^{2}}$ , this intends to a trivial group. That's where,my question, how could I know that ${\langle(a,b)\rangle: a^{m}=1=b^{n},(ab)^{k}=1}$ gives always $a,b,ab$ with order $m,n,k$ respectively? – A learner Jun 17 '21 at 15:24
  • The way to approach this problem is to construct groups with with the properties you want. The standard examples are triangle groups, which are groups of orientation-preserving actions of tilings of planes (spherical, Euclidean, or hyperbolic as appropriate) by triangles. (Wikipedia calls these "Von Dyke groups", using triangle groups for the non-orientation-preserving groups, but the papers and books I read called these groups triangle groups so I'm sticking with that...) – user1729 Jun 17 '21 at 15:24
  • By definition of a group presentation with $,a^m=1,$ we know that the order of $,a,$ divides $,m,$ so $,a,$ could be trivial. The exact order of $,a,b,ab,$ depends on the exact values of $,m,n,k.$ – Somos Jun 17 '21 at 15:27
  • @Somos so, you want to say my question answer is not affirmative. If not, can you give such proof with that particular $m,n,k$? – A learner Jun 17 '21 at 15:30
  • @user1729 please provide tha answer in answer section. I will be very grateful. – A learner Jun 17 '21 at 15:33

2 Answers2

6

Here is sketch of a completely different proof. Note that we need to assume that $m,n,k>1$.

Let $K$ be any field containing primitive $i$th roots of $1$ for $i=2m,2n,2k$. Then the group ${\rm PSL}(2,K)$ contains elements $a,b$ of order $m$ and $n$ such that $ab$ has order $k$. (So by choosing $K$ to be a finite field, we get finite examples.)

The construction makes use of the fact that the order of an element of ${\rm SL}(2,K)$ with distinct eigenvalues is determined by its trace.

So we can take $A= \left(\begin{array}{cc}\lambda&1\\0&\lambda^{-1}\end{array}\right)$ of order $2m$, $B = \left(\begin{array}{cc}\mu&0\\t&\mu^{-1}\end{array}\right)$ of order $2n$, where we choose $t \in K$ such that the trace of $AB$ is equal to $\nu + \nu^{-1}$, where $\nu$ has order $2k$ in $K$. Then $AB$ has order $2k$, and we define $a$ and $b$ to be the images of $A$ and $B$ in ${\rm PSL}(2,K)$.

(I have a feeling that this construction has been given here before.)

Example: $m=3$, $n=4$, $k=3$. We choose $\lambda$ and $\mu$ to be primitive $2m$th and $2n$th roots of unity, so 6th and 8th in this case.

Let $\nu$ be a primitive $2k$th root of unity (so 6th in this example).

The trace of $AB$ is $t+\lambda\mu + \lambda^{-1}\mu^{-1}$, so choose $t = \nu + \nu^{-1} - \lambda\mu - \lambda^{-1}\mu^{-1}$.

You can always choose $K = {\mathbb C}$. If you want a finite example, then choose $K = {\mathbb F}_q$, where $q-1$ is divisible by $2m$, $2n$ and $2k$. So in this case you could choose $q=25$, or $49$, or $73$, etc.

Derek Holt
  • 96,726
2

The way to approach this problem is to construct groups with with the properties you want. The standard examples are triangle groups*, which are groups of orientation-preserving actions of tilings of planes (spherical, Euclidean, or hyperbolic as appropriate) by triangles.

For example, if you want elements $a$ of order $2$ and $b$ of order $3$ such that $ab$ has order $6$ then the group of orientation-preserving symmetries of the Euclidean plane by triangle with angles $\pi/2$, $\pi/3$ and $\pi/6$ is what you want:

Wiki tiling

Here, $\pi/2+\pi/3+\pi/6=\pi$, which is what you're after for tiling the Euclidean plane (angles adding up to $\pi$ radians, so $180$ degrees). Therefore, dividing the above by $\pi$, we see that if your numbers $m,n,k$ are such that $1/n+1/m+1/k=1$ then there is a tiling of the Euclidean plan with triangles which has the properties you want. We can generalise: if $1/n+1/m+1/k<1$ then the tiling is of the hyperbolic plane, while if $1/n+1/m+1/k>1$ then the tiling is of the sphere. See the Wikipedia link above for more pictures/examples of these tilings.

*Wikipedia calls these "von Dyck groups", using triangle groups for the non-orientation-preserving groups, but the papers and books I read called these groups triangle groups so I'm sticking with that...!

user1729
  • 32,369
  • So, if I took $\frac{m\pi}{3},\frac{n\pi}{3},\frac{k\pi}{3}$ , it would work . Right? So, my question answer is affirmative. – A learner Jun 17 '21 at 16:01
  • @Alearner The sum of those three numbers is $(m+n+k)\pi\neq\pi$, so that does not work. To be clear: yes the answer to your question is affirmative. However, the answer is hard. – user1729 Jun 17 '21 at 16:04
  • No, you are taking me wrong ,, I am saying, $\frac{1}{m}\frac{m\pi}{3}+\frac{1}{n}\frac{n\pi}{3}+\frac{1}{k}\frac{k\pi}{3}$. Nice concept, I didn't see that thing before. I will work on it.Thanks, by the way :) – A learner Jun 17 '21 at 16:07
  • @Alearner I'm not sure what you mean, but there are precisely $3$ sets of integers $m,n,k$ with $1/m+1/n+1/k=1$. So for basically all sets of integers you are looking at tilings of the hyperbolic plane or the sphere. – user1729 Jun 17 '21 at 16:11
  • Oo, sorry , it's ok ,now. Thanks again :) – A learner Jun 17 '21 at 16:15