Edit: Is it possible to non-recursively define a (possibly infinite) binary tree? Here is my attempt with partial functions $L$ and $R$ on the set of natural numbers $N$ with nodes numbered by $N$
- $\forall x,y:[L(x)=y \implies x,y\in N]$
- $\forall x,y\in N: [L(x)=L(y) \implies x=y]$
- $\forall x,y:[R(x)=y \implies x,y\in N]$
- $\forall x,y\in N: [R(x)=R(y) \implies x=y]$
- $n\in N$
$\forall x\in N: [L(x)\neq n \land R(x)\neq n]$
$\forall x,y\in N: [L(x)=y => \forall z\in N: R(z)\neq y]\space\space$ (Edit)
- $\forall x,y\in N: [R(x)=y => \forall z\in N: L(z)\neq y]\space\space$ (Edit)
where
$L(x)=y$ means $y$ is the left child of node $x$
$R(x)=y$ means $y$ is the right child of node $x$
$n$ is the root node
No, these axioms are far too weak. For instance, you could have $N=\{n,a,b\}$ where $a=L(b)$ and $b=L(a)$.
Note that more generally, it will be impossible to give any first-order axiomatization of binary trees over the language $\{L,R,n\}$. The statement that every element is obtained from $n$ by finitely many applications of $L$ and $R$ cannot be expressed in first-order logic (e.g., by compactness).