I'm having trouble understanding why we need to add constants to a language in order to prove something when using the compactness theorem, in particular this:
Let L be a language with just a 2 place-symbol < for "less than". Show that there is no L-theory T such that the models of T are precisely the well-ordered sets.
The solution I have for this is: make a new language L' = L $\cup$ {$c_n | n \in \mathcal{N}$} and assume there is such theory T. Then T $\cup$ {$c_n > c_{n+1} | n \in \mathcal{N}$} is inconsistent so there is a finite subtheory that's inconsistent (by compactness theorem). Then there is a contradiction because this theory is true in $\mathcal{N}$.
Now my question is: why is it needed to add such new constants? I would not add them and say: $\phi_n = \exists x_1, \dots, x_n : x_1 >\dots > x_n$ and then take the theory T $\cup$ {$\phi_n | n \in \mathcal{N}$}, but this is apparently not correct.