For example, the structure formed as a disjoint sum of $\langle\mathbb{N},0,S,+,\cdot\rangle$ and $\langle\mathbb{Z},+\rangle$ is a model of PA$^{(+)U}\cup G^{U'}$ that has an obvious expansion to a model of PAPR$^{(+)U}\cup G^{U'}$, and the set of prime integers is definable in the expanded model but not in the original one. (Every subset of $\mathbb{Z}$ definable in the disjoint sum of $\langle\mathbb{N},0,S+,\cdot\rangle$ and $\langle\mathbb{Z},+\rangle$ is already definable in $\langle\mathbb{Z},+\rangle$—that is just an elementary fact about disjoint sums that can be proved by a straightforward induction. By the elimination of quantifiers, the primes are not definable in $\langle\mathbb{Z},+\rangle$, (Presburger), and they are therefore not definable in the disjoint sum of $\langle\mathbb{N},0,S+,\cdot\rangle$ and $\langle\mathbb{Z},+\rangle$. But the set of primes of $\mathbb{Z}$ is easily seen to be definable in the expanded structure—since the prime natural numbers are definable in $\langle\mathbb{N},0,S+,\cdot\rangle$—by using a primitive recursive definition of an injection from $\mathbb{N}$ into $\mathbb{Z}$.) Thus, the expansion is not an expansion by definitions.

-- ShaughanLavine - 16 Mar 2005