Ternary algebra satisfying some identities is a join-semilattice

63 Views Asked by At

A join-semilattice with greatest element is an algebra $(S,\vee, 1)$ of type $(2,0)$ such that $\vee$ is idempotent, commutative, and associative, and $a\vee 1=1$ for all $a\in A$.

Now, let $(A,m,1)$ be an algebra of type $(3,0)$ such that $m$ satisfies the following three identities:

(1) $m(x,y,x)=x$

(2) $m(x,x,y)=m(y,y,x)$

(3) $m[\, m(x,x,y)\,,\, m(x,x,y)\,,\, z\,]=m[\,x\,,\,x\,,\,m(y,y,z)\,]$

for all $x,y,z\in A$.

Define $x\vee y:= m(x,x,y)$ for $x,y\in A$. The claim is that $(A,\vee, 1)$ is a join-semilattice with greatest element.

Certainly $(A, \vee, 1)$ is of type $(2,0)$. Idempotence, commutativity, and associativity of $\vee$ follows from (1), (2), and (3), respectively.

However, I can't figure out why $x\vee 1=1$. Obviously I just need the right combination of (1), (2), and (3), but I'm not seeing it.

The best I can do right now is: \begin{align*}x\vee 1 = m(x,x,1) &= m\bigl[\,m(x,x,x)\,,\, m(x,x,x)\,,\, 1\,\bigr]\tag{1}\\ &=m\bigl[\,x\,,\,x\,,\,m(x,x,1)\,\bigr]\tag{3}\\ &=x\vee m(x,x,1)\,. \end{align*}

For reference, the claim is from this paper: last paragraph on p.129.

Thank you!

1

There are 1 best solutions below

0
On BEST ANSWER

As it was indicated in the comments, I'll post an answer to this question.


The claim is false. This will not hold unless you have some equation relating $1$ to $m$, otherwise there is no reason that this constant will be interpreted as the top element in the semilattice. More precisely, if $(A,\vee)$ (with $\vee$ defined as above) has at least two elements, you may interpret $1$ as any element different from the maximum (if it exists).

For an example, take $A:=\mathbb{Z}$ and $m(x,y,z):=\max\{\min\{x,y\},z\}$. It can be easily verified that this choice satisfies the three identities, and certainly $x\vee y= \max\{x,y\}$ is a semilattice operation. But $A$ is not bounded so actually there is no way of interpreting $1$ as the top (or, by the same token, bottom) element of the semilattice.

Finally, notice that in this paper, there is one more identity $m(x,x,1)=1 $ (which is exactly what you need). It seems fairly probable that they meant to include it in the list and simply forgot.