2

Let $T$ denote an algebraic theory. Then given a free $T$-algebra $F(k)$, the inclusion of generators map $\eta : k \rightarrow U(F(k))$ is usually injective, in practice. It doesn't have to be, though; consider the case where $T$ proves the identity $x \equiv y$. Then for $k \geq 2$, we have that $\eta : k \rightarrow U(F(k))$ is non-injective.

Question. What must we require of an algebraic theory $T$ for it to follow that for every cardinal number $k$, the inclusion of generators map $k \rightarrow U(F(k))$ is injective?

goblin GONE
  • 67,744

1 Answers1

1

The inclusion of generators map $i_\kappa:\kappa\rightarrow U(F(\kappa))$ is injective if and only if $T$ does not imply the identity $x = y$.

You've noted one direction in your question. For the converse, note that if $T$ does not imply $x = y$, then $T$ has at least one model, $A$, with $|A|\geq 2$. Pick $a,b\in A$ with $a\neq b$. Then for any cardinal $\kappa$, for any $\alpha, \beta\in \kappa$ with $\alpha\neq \beta$, there is a map $f:F(\kappa)\rightarrow A$ such that for $U(f):U(F(\kappa))\rightarrow U(A)$, we have $i_\kappa(\alpha)\mapsto a$ and $i_\kappa(\beta)\mapsto b$. Since $a\neq b$, $i_\kappa(\alpha)\neq i_\kappa(\beta)$, so $i_\kappa$ is injective.

Alex Kruckman
  • 76,357
  • I think there are some typos; in particular $f$ does not really feature in the argument; also, I think those $\mapsto$ symbols should be $=$. – goblin GONE Jul 07 '14 at 16:26
  • By $i_\kappa(\alpha) \mapsto a$, I mean that $i_\kappa(\alpha)$ maps to $a$ under $U(f)$, i.e. $U(f)(i_\kappa(\alpha)) = a$. The role of $f$ is just that if $i_\kappa(\alpha) = i_\kappa(\beta)$, their images under any function must be equal, but $f$ is chosen (by the universal property of the free algebra) to map them to unequal elements. – Alex Kruckman Jul 07 '14 at 16:47