Equational Theories

14 Equation 1516

In this chapter we study magmas that obey equation 1516,

\begin{equation} \label{1516} x = (y \diamond y) \diamond (x \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

\[ L_{Sy} L_x^2 y = x. \]

We begin by studying a greedily constructed translation invariant model

\begin{equation} \label{xy-1516} x \diamond y = x+f(y-x) \end{equation}
2

on the carrier \(\mathbb {Z}\) with some function \(f:\mathbb {Z}\to \mathbb {Z}\) with \(f(0)=0\). This ensures that \(Sx = x\). If \(y = x+h\), then \(L_x y = x + f(h)\), \(L_x^2 y = x + f^2(h)\), and \(L_y L_x^2 y = y + f(f^2(h)-h)\), so the law 1 simplifies to

\begin{equation} \label{1516-f} f(f^2(h)-h) = -h. \end{equation}
3

Thus, if we let \(E = \{ (h, f(h))\} \) be the graph of \(f\), then we have the following property: if \((a,b), (b,c) \in E\), then \((c-a,-a) \in E\). This helps motivate the following definition.

Definition 14.1 1516 seed
#

A 1516 seed is a finite collection \(E\) of pairs \((a,b)\) with \(a,b \in \mathbb {Z}\) obeying the following axioms:

  • Axiom 1: \((0,0) \in E\).

  • Axiom 2: If \((a,b) \in E\) and \(a \neq 0\), then \(b \neq 0, -a\).

  • Axiom 3: If \((a,b), (a,b') \in E\), then \(b=b'\).

  • Axiom 4: If \((a,b), (b,c) \in E\), then \((c-a,-a) \in E\).

  • Axiom 5: If \((b,a), (b',a), (-b, d), (-b',d') \in E\) with \(b \neq b'\), then \(b+d \neq d', b'+d'\).

An extension of a 1516 seed \(E\) is a 1516 seed \(E'\) that contains \(E\).

This definition has an extension property:

Lemma 14.2 1516 extension

Let \(E\) be a 1516 seed, and let \(a_0 \in \mathbb {Z}\). Then there exists an extension \(E'\) of \(E\) that contains \((a_0,c_0)\) for some \(c_0\).

Proof

We may assume that \(E\) does not already contain any pair of the form \((a_0,c)\), since we are done otherwise. By Axiom 1 implies that \(a_0 \neq 0\). Let \(b_1,\dots ,b_n\) denote all the integers \(b_i\) such that \((b_i,a) \in E\), then \(n \geq 0\) is finite, and by Axiom 2 all the \(b_i\) are non-zero and not equal to \(-a\). Let \(c_0\) be a sufficiently large integer to be chosen later. We then add the pairs \((a_0,c_0)\) and \((c-b_i,-b_i)\) to \(E\) for all \(i\). Furthermore, if \(i\) is such that \((-b_i,d_i) \in E\) for some (necessarily unique and non-zero) \(d_i\), we also add \((d_i+b+i-c_0, b_i-c_0)\) to \(E\). Let \(E'\) be the resulting set of pairs. Axioms 4, 5 for \(E\) ensure that (for \(c_0\) large enough) the addition of these pairs do not cause a violation of Axioms 2 or 3 for \(E'\), and of course Axiom 1 for \(E'\) will also be retained. As for Axiom 4, one can check that the only new pairs of pairs \((',b), (b,c)\) that would trigger these axioms either take the form \((b_i,a_0), (a_0,c_0)\) or \((c_0-b_i,-b_i), (-b_i,d_i)\), and in either case we see from construction that Axiom 4 remains in effect for \(E'\). Finally, the new quadruples of pairs \((b,a), (b',a), (-b, d), (-b',d') \in E'\) only arise when either \((-b,d)\), \((-b',d')\) is equal to \((a_0,c_0)\) and the other three pairs in the quadruple were already in \(E\), and for \(c_0\) large enough we see that Axiom 5 remains valid.

For technical reasons (which will be helpful later when we expand the magma) we also give a variant that will ensure a useful “double surjectivity” property:

Lemma 14.3 1516 extension variant

Let \(E\) be a 1516 seed, and let \(h \in \mathbb {Z}\) be non-zero. Then there exists an extension \(E'\) of \(E\) that contains \((a,a+h)\), \((a',a'+h)\), \((a'',a''+h)\) for some distinct \(a,a',a''\).

Proof

If we choose \(a\) sufficiently large, and set \(a' = 2a\) and \(a''=3a\) (say), the claim simply follows by adding \((a,a+h)\), \((a',a'+h)\) and \((a'',a''+h)\) to \(E\) and verifying that none of the axioms are violated.

Corollary 14.4 Base magma

There exists a 1516 magma with carrier \(\mathbb {Z}\) with the property that \(Sa=a\) for all \(a\), and such that for any distinct \(a,b \in \mathbb {Z}\), the equation \(R_a c = b\) has at least three solutions \(c\). Furthermore, for each \(a \in \mathbb {Z}\), there exists \(b \neq a\) such that \(L_a R_a b = b\).

Note from 1 and the hypothesis \(Sa=a\) that \(R_a c = a\) if and only if \(c=a\). Thus the requirement that \(a,b\) be distinct is necessary.

Proof

By Lemma 14.2, Lemma 14.3 and the greedy algorithm starting with the seed consisting of \((0,0)\), \((-1,2)\), \((3,1)\), we obtain a graph \(\{ (a,f(a)): a \in \mathbb {Z}\} \) of a function \(f:\mathbb {Z}\to \mathbb {Z}\) with \(f(0)=0\), \(f(-1)=2\), \(f(3)=1\) and the property that if \(f(a)=b\) and \(f(b)=c\), then \(f(c-a)=-a\), and also with the property that for every non-zero \(h\) there are distinct \(a_h, a'_h, a''_h\) with \(f(a_h)=a_h+h\), \(f(a'_h) = a'_h+h\), and \(f(a''_h) = a''_h+h\). We then have 3 holds. If we then define the magma operation \(\diamond \) by 2, we obtain 1. Also, from construction we see that \(R_d (d - a_h) = R_d (d - a'_h) = R_d (d - a''_h) = d+h\) for any \(d,h\), giving the second claim. Finally, from constructio we have \(R_a (a+1) = a+1+f(-1) = a+3\), so \(L_a R_a (a+1) = a + f(3) = a+1\), giving the final claim.

Now we construct a more complex 1516 magma, whose carrier \(G\) is \(\mathbb {Z}\cup G'\), where

\[ G' := \{ (a,c,n) \in \mathbb {Z}\times \mathbb {Z}\times \mathbb {N}: a \neq c\} . \]

The first component \(\mathbb {Z}\) will represent the squares, and the second component \(G'\) will represent the non-squares, and the third component is a technical label needed to ensure a certain infinite surjectivity property. The 1516 magma constructed in Corollary 14.4 will be the restriction of \(G\) to \(\mathbb {Z}\); thus \(a \diamond b\) is already defined in \(G\) for \(a,b \in \mathbb {Z}\), but the rest of the multiplication table is not currently defined. By construction, \(S\) is already defined and equal to the identity on \(\mathbb {Z}\), and we have the 1516 equation

\[ L_{Sa} L_b L_b a = b \]

for \(a,b \in \mathbb {Z}\), with the left multiplication operators \(L_b\) currently only defined as maps from \(\mathbb {Z}\) to \(\mathbb {Z}\). Among other things, this means that \(L_a = L_{Sa}\) is surjective as a map from \(\mathbb {Z}\) to \(\mathbb {Z}\) for any \(a \in \mathbb {Z}\).

We extend the squaring map \(S\) to all of \(G\) by declaring \(S(a,b,n_1,n_2) := a\), thus \(S\) maps \(G\) to \(\mathbb {Z}\). To create a 1516 magma structure on all of \(G\), we need to extend the left multiplication operators \(L_b\), \(b \in \mathbb {Z}\) as maps from \(G\) to \(G\), and also introduce additional maps \(L_x: G \to G\) for \(x \in G'\), obeying the following axioms:

  • Axiom A: For any \(x \in G'\), \(L_x x = Sx\).

  • Axiom B: For any \(x \in G'\) and \(b \in \mathbb {Z}\), \(L_{Sx} L_b L_b x = b\).

  • Axiom C: For any \(x \in G\) and \(y \in G'\), \(L_{Sx} L_y L_y x = y\).

We first address Axiom B, which purely concerns how to extend the \(L_b\) operators for \(b \in \mathbb {Z}\), and also impose an additional technical “infinitely surjective” requirement that will help us satisfy Axiom C later.

Proposition 14.5 Obtaining Axiom B

There exists a way to extend \(L_b: \mathbb {Z}\to \mathbb {Z}\) to \(L_b: G \to G\) for all \(b \in \mathbb {Z}\), in such a way that Axiom B holds, and furthermore for each \(b \in \mathbb {Z}\) and \(x \in G'\), the set \(\{ y \in G': L_b y = x \} \) is infinite. Also we can ensure that \(L_b x \neq x\) for any \(b \in \mathbb {Z}\) and \(x \in G'\).

Proof

Let \(y = (a,c,n)\) be a “non-square” in \(G'\), thus \(a \neq c\) and \(Sy = a\). We first define the operation \(L_a\) on \(y\) as follows:

  • If \(n \neq 0\), we set \(L_a y := (a,c,0)\).

  • If \(n=0\), we set \(L_a y = a\).

In particular, we obtain the special case \(L_{Sy} L_a L_a y = a\) of Axiom B. We also see that the the preimage \(\{ z \in G': L_a z = y \} \) is already infinite if \(n = 0\), but this has not yet been established in the \(n \neq 0\) case.

We still need to define \(L_c y\) for other \(c \in \mathbb {Z}\). We claim that for each \(b \in \mathbb {Z}\) we can find \(c_{y,b} \in \mathbb {Z}\), which are distinct as \(b\) varies, such that \(L_a L_{c_{y,b}} b = c_{y,b}\). For \(b=a\), we can write \(L_a L_{c_{y,b}} b = L_a R_a c_{y,b}\) and invoke the final property of Corollary 14.4. For \(b \neq a\), we apply Corollary 14.4 to find \(c_{y,b} \in \mathbb {Z}\) distinct from \(b, c_{y,a}\) such that \(R_a c_{y,b} = b\); note that \(c_{y,b}\) is also distinct from \(a\) since \(R_a a = a \neq b\). Clearly the \(c_{y,b}\) are distinct as \(b\) varies. We now have \(L_a L_{c_{y,b}} b = L_{Sa} L_{c_{y,b}} L_{c_{y,b}} a = c_{y,b}\) as desired.

So if we define \(L_{c_{y,b}} y := b\) for all \(b \in \mathbb {Z}\), then we have \(L_{Sy} L_{c'} L_{c'} y = c'\) whenever \(c'\) is of the form \(c' = c_{y,b}\) for some \(b \in \mathbb {Z}\). But we still have \(L_{c'} y\) undefined if \(c'\) is not of this form or equal to \(a\), in particular, \(L_c y\) is currently not defined.

We also note that with our construction so far, that \(L_c y \neq c\) whenever \(y \in G'\) and \(c \in \mathbb {Z}\) is distinct from \(Sy\). This proeprty will be preserved in the construction below (because all future assignments of \(L_c y\) will take values in \(G'\) rather than \(\mathbb {Z}\)).

Now we greedily fill in the rest of the \(L_{c'}\) functions. Suppose at some stage of this process, we have a non-square element \(y = (a,c,n)\) for which \(L_{c'} y\) is not yet defined for some \(c'\), then \(c' \neq a\). We can use the surjectivity of \(L_a: \mathbb {Z}\to \mathbb {Z}\) to write \(c' = L_a b\) for some \(b \in \mathbb {Z}\), then set \(L_{c'} y = b\); then we have \(L_{Sy} L_{c'} L_{c'} y = c'\).

Next, suppose we want to write \(y\) as \(L_{c'} z\) for some \(z\) where \(L_{c'}\) is not currently defined, in such a way that \(L_{Sz} L_{c'} L_{c'} z = c'\). There are several cases:

  • Case 1: \(L_{c'} y = w\) for some \(w \in G'\). By construction, \(c_{w,c'}\) is distinct from \(c'\) and \(L_{c_{w,c'}} w = c'\). Set \(z = (c_{w,c'}, c', n', 0)\) where \(n'\) is large enough that \(L_{c'} z\) has not yet been assigned. Then set \(L_{c'} z = y\), so that we have \(L_{Sz} L_{c'} L_{c'} z = c'\) as required.

  • Case 2: \(c' \neq Sy\) and \(L_{c'} y = b\) for some \(b \in \mathbb {Z}\). By the previous property, \(c' \neq b\). By Corollary 14.4, we can find \(a'\) such that \(R_b a' = c'\) and \(a' \neq c'\). Then set \(z = (a', c', n', 0)\) where \(n'\) is large enough that \(L_{c'} z\) has not yet been assigned. Then set \(L_{c'} z = y\), so that we have \(L_{Sz} L_{c'} L_{c'} z = c'\) as required.

  • Case 3: \(c' \neq Sy\) and \(L_{c'} y\) is currently undefined. Use the previous procedure to set \(L_{c'} y\) equal to some \(b \in \mathbb {Z}\), then apply the Case 2 analysis.

  • Case 4: \(c' = Sy\). We can assume \(n=0\), since otherwise we are in Case 1 by construction. But then we do not need to locate any \(z\) because, as previously mentioned, the set \(\{ z \in G': L_{c'} z = y \} \) is already infinite.

Iterating these procedures over a well-ordering of \(G' \times \mathbb {Z}\times \mathbb {N}\), we can show that for any finite initial segment in this well-ordering, we can create a partially defined extension of all the \(L_b\), such that if \((x,c',m)\) is in this initial segment, then \(L_{Sx} L_{c'} L_{c'} x\) is well-defined and equal to \(x\), and \(\{ y \in G': L_b y = x \} \) has at least \(m\) elements. Running this iteration over the entire well-ordering, we obtain the claim.

Finally, we construct the remaining \(L_x\).

Proposition 14.6 Obtaining Axioms A, C

One can find maps \(L_x: G \to G\) for each “non-square” \(x \in G'\), such that Axioms A, C hold.

Proof

We can work with a single \(x \in G'\). Our task is to find a function \(L_x\) for which

\begin{equation} \label{axioma-again} L_x x = Sx \end{equation}
4

and

\begin{equation} \label{axiomb-again} L_{Sy} L_x L_x y = x \end{equation}
5

for all \(y \in G\) (note that \(L_{Sy}\) is already fully constructed).

Define a seed to be an injective partial function \(L_x\) defined on finitely many values and obeying 4, as well as 5 whenever \(L_x L_x y\) is defined. If there is a \(y \in G\) for which \(L_x y\) is currently undefined, then by hypothesis \(y\) is equal to \(L_x z\) for at most one \(z\). If such a \(z\) exists, we set \(L_x y\) to be an element of \(\{ w: L_{Sz} w = x \} \) that is not already in the domain or range of \(L_x\); if no such \(z\) exists, we set \(L_x y\) to be arbitrary element of \(G\) not already in the domain or range. In either case, we see that the seed property is preserved. Starting from the seed in which \(L_x\) is only defined on \(x\) and maps it to \(Sx\), we obtain the claim.

Corollary 14.7

There exists a 1516 magma that does not obey the 255 equation

\[ x = ((x \diamond x) \diamond x) \diamond x \]

.

Proof

We perform the above construction, but with one refinement. Pick some \(x \in G'\). Then after the constructions in Proposition 14.5, \(L_{Sx} x\) is defined and equal to some element \(y\) distinct from \(x\). If \(y \in \mathbb {Z}\) then \(L_y x \neq x\) by construction and we are done, so suppose \(y \in G'\). It is then a routine matter to modify the construction in Proposition 14.6 to ensure that \(L_y x \neq x\), by adding one more element to the seed for \(L_y\) to set \(L_y x\) to some arbitrary value not equal to \(x, y, Sy\).