# A question about (unicity of certain cycles in a Cayley graph of a) symmetric group

Let $S=\{(1,2),(1,2,3,\ldots,n),(1,2,3,\ldots,n)^{-1}=(1,n\ldots,2)\}$ be a subset of the symmetric group $S_n$. We know that $(1,2,\ldots,n)(1,2)=(2,3,\ldots,n)$, and thus $$[(1,2,\ldots,n)(1,2)]^{n-1}=(1,2,\ldots,n)\overbrace{(1,2)\cdots(1,2,\ldots,n)}^{2n-2}(1,2)=(1).$$ We want to know whether or not there exists another sequence of elements $a_1,a_2,\ldots,a_{2n-4}\in S$ such that $$(12\ldots n)a_{2n-4}a_{2n-5}\cdots a_2a_1(12)=(1),$$ where $a_{i+1}\neq a_i^{-1}$ for $i=0,1,2,\ldots,2n-4$ (putting $a_0=(12)$, $a_{2n-3}=(1,2,\ldots,n)$).

Equivalently, I want to ask if, in the cubic Cayley graph $Cay(S_n,S)$, there is a unique cycle of length $2(n-1)$ passing through $(1)$, $(1,2,\ldots,n)$ and $(1,2)$.

• should $S$ be the subgroup generated by the three permutations or just the set of the three given permutations? – Amir Sagiv Apr 20 '16 at 16:13
• you understand $S=S_n$? – Mikhail Ivanov Apr 20 '16 at 16:40
• @AmirSagiv: If it were meant to be the subgroup generated, then it would be equal to all of $S_n$ (in fact, the first two already generate $S_n$). So surely not: he's asking about expressing the identity as a word in the elements of the subset $S$ with specific constraints. – Arturo Magidin Apr 20 '16 at 17:06
• In other words, you are asking if, in the cubic Cayley graph on $S_n$ with the generators you have given, there is another path of length $2n-4$ between the identity and $(1,n,n-1,\ldots,3)$ besides the one you give... – verret Apr 20 '16 at 20:10
• For small values of n (n < 10), a computer program should be able to arrive at the answer using a brute force enumeration. Have you tried such? Gerhard "Sometimes Small Informs The Large" Paseman, 2016.04.20. – Gerhard Paseman Apr 21 '16 at 5:06

The smallest $n$ for which there exist sequences as asked for is $n = 7$:

• $(1,2,3,4,5,6,7) \cdot (1,2) \cdot (1,7,6,5,4,3,2) \cdot (1,2) \cdot (1,2,3,4,5,6,7) \cdot (1,2) \cdot$ $(1,7,6,5,4,3,2) \cdot (1,2) \cdot (1,2,3,4,5,6,7) \cdot (1,2) \cdot (1,7,6,5,4,3,2) \cdot (1,2) = ()$, and

• $(1,2,3,4,5,6,7) \cdot (1,2,3,4,5,6,7) \cdot (1,2) \cdot (1,7,6,5,4,3,2) \cdot (1,7,6,5,4,3,2) \cdot$ $(1,2) \cdot (1,2,3,4,5,6,7) \cdot (1,2,3,4,5,6,7) \cdot (1,2) \cdot (1,7,6,5,4,3,2) \cdot$ $(1,7,6,5,4,3,2) \cdot (1,2) = ()$.

For $n = 8$ there is no such sequence other than the trivial one mentioned in the question, for $n = 9$ there is one, for $n = 10$ there are $18$, for $n = 11$ there are $5$ and for $n = 12$ there are $104$ such sequences. This has been found with the following GAP function:

SearchXueyiSequences := function ( n )

local  sequences, search, S;

search := function ( sequence, a )

local  b;

if Length(sequence) = 2*n-3 then
if sequence[2*n-3] <> (1,2) and Product(sequence)*(1,2) = () then
fi;
return;
fi;
for b in Difference(S,[a^-1]) do
search(Concatenation(sequence,[a]),b);
od;
end;

S := Set(GeneratorsAndInverses(SymmetricGroup(n)));
sequences := [];
search([],S[2]);
sequences := Set(sequences);
sequences := sequences{[2..Length(sequences)]}; # exclude trivial sequence
return sequences;
end;


For the case $n=3$, take $a_1 = (1,2)$ and $a_2 = (1,3,2)$ instead of $a_1 = (1,2,3)$ and $a_2 = (1,2)$.

A major motivation for posting this answer is to give Xueyi Huang an opportunity to exclude this case in one way or another if it is not really relevant.

Further edit on April 21: As anticipated, Xueyi Huang has made a minor modification to the earlier form of the question and this answer is not a valid answer to the current form of the question.

• When $n=3$ and $n=4$, such a sequence is unique by simple computation. I want to know whether the conclusion is true for general $n$. – Xueyi Huang Apr 21 '16 at 2:43
• You cannot take $a_1=(1,2)$ and $a_2=(1,3,2)$ instead of $a_1=(1,2,3)$ and $a_2=(1,2)$ since they don't satisfy the condition $a_{i+1}\neq a_i^{-1}$. – Xueyi Huang Apr 21 '16 at 3:07