Equational Theories

13 Equation 1323

In this chapter we study magmas that obey equation 1323,

\begin{equation} \label{1323} x = y \diamond (((y \diamond y) \diamond x) \diamond y) \end{equation}
1

for all \(x,y\). Using the squaring operator \(Sy := y \diamond y\) and the left and right multiplication operators \(L_y x := y \diamond x\) and \(R_y x = x \diamond y\), this law can be written as

\[ L_y R_y L_{Sy} x = x. \]

Among other things, this implies that \(L_{Sy}\) is injective and \(L_y\) is surjective, hence \(L_{Sy}\) is invertible. It is convenient to impose an additional axiom

\begin{equation} \label{lr} L_{Sy} R_{Sy} = 1 \end{equation}
2

that is to say \(R_{Sy}\) is the inverse of \(L_{Sy}\), then the law 1 simplifies to

\begin{equation} \label{lr-simp} L_{y} R_{y} = R_{Sy}. \end{equation}
3

So now we would like to construct a magma satisfying 2 and 3. Let \(G\) be a countably infinite abelian group of order \(2\), and work on a carrier \(G \uplus N\), where \(N\) is the set of pairs \((x,a)\) with \(x \in \mathbb {Q}^\times \) and \(a \in G \backslash \{ 0\} \). For each \(a \in G \backslash \{ 0\} \), we let \(\phi _a: G \to \mathbb {Q}^\times \) be a bijection such that \(\phi _a(0) = 1\) and \(\phi _a(a+b) = -1/\phi _a(b)\) for all \(b \in G\), so in particular \(\phi _a(a)=-1\); such a bijection can be easily constructed from the axiom of choice and a greedy algorithm, defining \(\phi _a\) one pair \(\{ b,a+b\} \) at a time. We then partially define an operation \(\diamond \) on \((G \times N) \times (G \times N)\) by setting

\begin{equation} \label{op-0} a \diamond b = a+b \end{equation}
4

for \(a,b \in G\), and

\begin{equation} \label{op-1} (x,a) \diamond b = (\phi _a(b) x, a) \end{equation}
5

\begin{equation} \label{op-2} b \diamond (x,a) = (x / \phi _a(b), a) \end{equation}
6

and

\begin{equation} \label{op-3} (\phi _a(b) x,a) \diamond (x,a) = a+b \end{equation}
7

for \(x \in \mathbb {Q}^\times \), \(b \in G\), and \(a \in G \backslash \{ 0\} \). This defines \(\diamond \) except for \((x,a) \diamond (y,b)\) with \(x,y \in \mathbb {Q}^\times \) and \(a,b,0\) distinct.

With these rules, \(0\) is a unit, and the squaring operator is given by \(Sa = 0\) and \(S(x,a) = a\), so the set of squares is \(G\). If \(a \in G\) is a square number, we have

\[ L_a b = R_a b = a+b \]

and

\[ L_a (x,b) = (x/\phi _b(a), b); \quad R_a (x,b) = (x \phi _b(a), b) \]

so 3 is satisfied.

We can also verify some cases of 3. If \(y\) is a square, then the claim already follows from 2 since \(Sy=0\) is a unit. Otherwise, for \(y \in \mathbb {Q}^\times \) and \(b \in G \backslash \{ 0\} \), we have to show that

\[ L_{(y,b)} R_{(y,b)} a = R_b a = a+b \]

for \(a \in G\) and

\[ L_{(y,b)} R_{(y,b)} (x,a) = R_b (x,a) = (\phi _a(b) x, a) \]

for \(x \in \mathbb {Q}^\times \) and \(a \in G \backslash \{ 0\} \). In the first case, we have from 6 that

\[ R_{(y,b)} a = (y/\phi _a(b), b) \]

and then from 7 we have

\[ L_{(y,b)} R_{(y,b)} a = a+b \]

as required. In the second case, if \(a=b\), then by the bijective nature of \(\phi _a\), we can write \(x = \phi _a(c) y\) for some \(c \in G\), then from 7 we have

\[ R_{(y,b)} (x,a) = a+c \]

and then by 5

\[ L_{(y,b)} R_{(y,b)} (x,a) = (\phi _a(a+c) y, a) \]

but from construction we have \(\phi _a(a+c)=-1/\phi _a(c) = \phi _a(b)/\phi _a(c)\) and hence \(\phi _a(a+c) y = \phi _a(b) x\) as required.

It remains to handle the case when \(a,b\) are distinct elements of \(G \backslash \{ 0\} \). Here we will set \((x,a) \diamond (y,b)\) to some \((z,c)\) with \(c\) distinct from \(a,b,0\) to be determined. The axiom to verify is then

  • Axiom: If \((x,a) \diamond (y,b) = (z,c)\), then \((y,b) \diamond (z,c) = (\phi _a(b) x,a)\).

To do this we perform a greedy algorithm. Suppose that \((x,a) \diamond (y,b)\) is undefined for some \(x,y \in \mathbb {Q}^\times \) and distinct \(a,b \in G \backslash \{ 0\} \). We select a \(c \in G \backslash \{ 0\} \) that has not previously been used by the greedy algorithm, and set

\[ (x,a) \diamond (y,b) = (z,c) \]

for some arbitrary \(z \in \mathbb {Q}^\times \) (e.g., one could take \(z=1\) if desired). To verify the axiom, we then also need to make an infinite number of other assignments, namely

\[ (\phi _a(b)^n x,a) \diamond (\phi _b(c)^n y,b) = (\phi _c(a)^n z,c) \]
\[ (\phi _b(c)^n y,b) \diamond (\phi _c(a)^n z,c) = (\phi _a(b)^{n+1} x, a) \]
\[ (\phi _c(a)^n z,c) \diamond (\phi _a(b)^{n+1} x, a) = (\phi _b(c)^{n+1} y,b) \]

for all natural numbers \(n\). As the \(a,b,c\) are distinct, the \(\phi _a(b), \phi _b(c), \phi _c(a)\) are not equal to \(\pm 1\), and the members of this infinite sequence do not collide with each other. If we arrange matters so that \(\phi _b(c)\) involves primes in the numerator or denominator that do not appear in any previous \(\phi _a(b), \phi _b(c), \phi _c(a), x, y, z\) used by the greedy algorithm, or the current \(x,y,z\), then we also see that this infinite sequence does not collide with any previously assigned value of \(\diamond \) either. If we then add all these assignments to the multiplication table, we maintain the desired axiom. Iterating this a countable number of times, we obtain a magma that obeys 2 and 3, and hence 1323.

Corollary 13.1 1323 does not imply 2744

There exists a 1323 magma which does not obey the 2744 equation \(R_y L_{Sy} L_y x = x\).

Proof

It suffices to produce a model in which \(L_y\) is not injective. Pick distinct \(a, b, b', c\) with \(\phi _a(b)\), \(\phi _a(b')\) having no prime factors in common in the numerator or denominator. If we impose the conditions

\[ (\phi _a(b)^n,a) \diamond (\phi _b(c)^n,b) = (\phi _c(a)^n,c) \]
\[ (\phi _b(c)^n,b) \diamond (\phi _c(a)^n,c) = (\phi _a(b)^{n+1}, a) \]
\[ (\phi _c(a)^n,c) \diamond (\phi _a(b)^{n+1}, a) = (\phi _b(c)^{n+1},b) \]

and

\[ (\phi _a(b)^n,a) \diamond (\phi _{b'}(c)^n,b') = (\phi _c(a)^n,c) \]
\[ (\phi _b(c)^n,b) \diamond (\phi _c(a)^n,c) = (\phi _a(b')^{n+1}, a) \]
\[ (\phi _c(a)^n,c) \diamond (\phi _{a'}(b)^{n+1}, a) = (\phi _b(c)^{n+1},b') \]

then no collisions occur, and the required axiom is obeyed; on the other hand, as \(L_{(1,a)} (1,b) = L_{(1,a)} (1,b') = (1,c)\), we already have a violation of left injectivity. Completing this seed to a full magma, we obtain the claim.

A very similar construction gives

Corollary 13.2 1898 does not imply 2710
#

There exists a magma which obeys the 1898 law \(R_{Sy} L_y R_y x = x\) but not the 2710 law \(R_y R_{Sy} L_y = x\).