Prove that, for $n \geq 3$, the three-cycles generate the alternation group $A_n$

Proof: We multiply on the left by 3-cycles to "reduce" an even permutation $p$ to the identity, using induction on the number of indices fixed by a permutation. How the indices are numbered is irrelevant. If $p$ contains a $k$-cycle with $k \geq 3$, we may assume that it has the form $p=(123\dots k)\dots$ Multiplying on the left by $(321)$ gives $$p'= (321)(123 \dots k)\dots=(1)(2)(3\dots k)\dots$$ More fixed indices.

What do you think ?

  • $\begingroup$ How would you use your method to show that for instance $(12)(34)$ may be written as a product of 3-cycles? $\endgroup$ – paw88789 Aug 30 '14 at 22:39
  • $\begingroup$ Hi Carpe, But I thought you were multiplying by 3-cycles to get to the identity. $\endgroup$ – paw88789 Aug 30 '14 at 22:45
  • $\begingroup$ mmm.. Let me think about that. $\endgroup$ – Carpediem Aug 30 '14 at 22:48
  • $\begingroup$ Where have you used he fact that the permutation $p$ is even? If your argument worked, then why would it not "prove" that $S_{n}$ is generated by $3$-cycles (which it isn't)? $\endgroup$ – Geoff Robinson Aug 30 '14 at 23:43
  • $\begingroup$ Technical point: it is better to use Markdown syntax for **bold** font, because then the font matches the rest of text, unlike the font used by $\textrm{MathJax}$. $\endgroup$ – user147263 Aug 31 '14 at 4:29

A permutation is an element of $A_n$ if and only if it is a product of an even number of transpositions.

We first note that the three cycles do not generate more than $A_n$ since for distinct $i,j,$ and $k$, we have $(ij)(ik)=(ijk)$.

For the other inclusion, we note that $(ijk)=(ij)(ik)$ for distinct $i,j,$ and $k$. And for distinct $i,j,k,$ and $l$, we have that $(ij)(kl)=(ijk)(jkl)$. Thus the three cycles generate all products of an even number of transpositions.

  • $\begingroup$ Could you perhaps write you proof based on mine ? That would be more helpful. Thank you for you help $\endgroup$ – Carpediem Aug 30 '14 at 22:57
  • $\begingroup$ @Carpediem If you are looking at a cycle decomposition that contains only 2-cycles which looks like $(ij)(kl)w$ where $w$ is another permutation, then your original permutation is in the alternating group if and only if $w$ is in there. $\endgroup$ – Robert Wolfe Aug 30 '14 at 23:00
  • $\begingroup$ So what is wrong with what I wrote ? $\endgroup$ – Carpediem Aug 30 '14 at 23:04
  • $\begingroup$ Nothing is wrong with it, as long as what you're reducing has a cycle of length $k\geq 3$ out in front. If it doesn't, what do you do? If it has a cycle of length 2 out in front, you have a problem. But you don't if there's a cycle of length 2 after that by my previous comment. What if you have a 2-cycle then a k-cycle with $k\geq 3$? Still have a problem. You need to break up that 2nd cycle into a 2-cycle and another cycle. How do you do that? $\endgroup$ – Robert Wolfe Aug 30 '14 at 23:09
  • $\begingroup$ Ok, how about this? If you have a 2-cycle and a k-cycle out front, you can conjugate by the inverse of the 2-cycle. This new element will be in $A_n$ if and only if the original was. That will put the k-cycle out front. $\endgroup$ – Robert Wolfe Aug 30 '14 at 23:11

The idea of the proof is that we will show that any product of two transpositions is a product of 3-cycles ; and ,if so, then since every member of the An is a product of even number of transpositions, we are done.

To begin with lets us suppose that we have a product of transpositions a1 and a2 where both have a common number a which belongs to the set {1,2,.....}- say they both move a. Then we have a1 of the form (a b) and a2 of the form (a c) . In this case , we have

a1 a2 = (a b)(a,c)= (a,c,b) and we are done. Now suppose that a1 and a2 move different numbers .

a1 = (a b) a2 = (c d) Then

a1a2 = (a,b)(c,d) = ( d a c ) ( a b d ). And we are done.


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.