# Is this binary operation commutative?

In a set $X$ we define a binary operation $*$ such that

$$\forall x, y \in X,\ (x*y)*y=y*(y*x)=x.$$

Is $*$ commutative and why?

-
Do we also assume * is associative? – coffeemath Oct 21 '12 at 10:14
No, it's not associative. – Ezra Oct 21 '12 at 10:15
I guess it is not necessarily commutative. – Berci Oct 21 '12 at 10:18
I guess not. But is there a counterexample? – Ezra Oct 21 '12 at 10:19
Such a $(X,*)$ cannot be a subsemigroup of a group $G$, because then all $x\in X$ satisfies $x^2=1$, so $1=(xy)^2=x^2y^2 \implies yx=xy$. – Berci Oct 21 '12 at 10:48

Yes, it is commutative. Here is why : $$x * y = y*(y*(x*y)) = y * ((x * (x*y))*(x*y)) = y *x.$$ I first use the identity to multiply by $y$ on the left twice, and then I replace the second $y$ by $x*(x*y)$. Then we use the fact that $$(x * (x * y))*(x*y) = x$$ to remove the bigger chunk.

Hope that helps,

-
Simple idea : if you want to begin with $x*y$ and end up with $y*x$, you have to 'kill' an $x*y$, but the only way that can happen is if you can make it appear twice. So I wanted to put an extra $x*y$ in there, made it happen, and there you go. =D – Patrick Da Silva Oct 21 '12 at 11:06
You won. Nice one. – Berci Oct 21 '12 at 11:06
with $x=y*(y*x)$ the $x*y$ should be $(y*(y*x))*y$ essentially you proved $y*x=y*x$ – ratchet freak Oct 21 '12 at 13:38
@ratchet freak : Read it again. Are you absolutely sure that I am wrong? Note that I used that for $X = x * y$ the fact that $y * (y * X) = X$, i.e. that $y * (y * (x*y)) = x*y$. The extreme left of my equality is $x*y$, the extreme right is $y*x$ and in the middle I just keep using the identity with appropriate choices of $X$ and $Y$ to plug in. Just read it carefully, there's only three steps to find. =) – Patrick Da Silva Oct 21 '12 at 19:26

Ah, I think I got a counterexample:

Consider the free monoid $F_2$ on 2 letters $a,b$ (with empty word '$1$'), and consider its quotient $$X:=F_2/(w^2\sim 1)$$ for all words $w\in F_2$. Then, basically $X$ contains those words which doesn't have any substring of the form $ww$, and if such appears at a concatenation, they get just cancelled. For example $aba\cdot ab= a$.

Update: not good, as it is going to be a group, $xy=yx$ will follow from all $x^2=1$...

-
Do you think the quotient still satisfies the latter? – Patrick Da Silva Oct 21 '12 at 10:55
@PatrickDaSilva: It obviously does, since $y(yx)=(yy)x=x=x(yy)=(xy)y$ – tomasz Oct 21 '12 at 11:00
You guys should check my answer. – Patrick Da Silva Oct 21 '12 at 11:02
Hmm.. Something is still not OK:( it is a Group! – Berci Oct 21 '12 at 11:02
@tomasz : You assumed the operation is associative. – Patrick Da Silva Oct 21 '12 at 11:03

Here's a proof that uses the automated theorem prover Prover9.

As a human, I find the output of these things difficult to read, but I find they can help by (a) suggesting an important in-between step, and (b) give a proof (even if it is unreadable), which means the result we're trying to prove is actually true, and it is possible there is a proof with a number of steps that could be understood by humans. (Also, there is a "coolness factor" to automated proofs.)

Here's the input:

formulas(assumptions).

(x * y) * y = x.
y * (y * x) = x.

end_of_list.

formulas(goals).

x * y = y * x.

end_of_list.


and a trimmed version of the output:

============================== PROOF =================================

% Proof 1 at 0.01 (+ 0.00) seconds.
% Length of proof is 7.
% Level of proof is 3.
% Maximum clause weight is 7.
% Given clauses 4.

1 x * y = y * x # label(non_clause) # label(goal).  [goal].
2 (x * y) * y = x.  [assumption].
3 x * (x * y) = y.  [assumption].
4 c2 * c1 != c1 * c2.  [deny(1)].
5 x * (y * x) = y.  [para(3(a,1),2(a,1,1))].
7 x * y = y * x.  [para(2(a,1),5(a,1,2))].
8 \$F.  [resolve(7,a,4,a)].

============================== end of proof ==========================

-