I raise following question after reading this post.
Is it possible in the language $L:= \{0,1,+,\cdot\}$ to write sentences for which a model will necessarily contain a copy of $\mathbb Q$ and a non algebraic element?
If yes, what is this set of sentences? If not can you provide a proof?
Thanks.
I added the word necessarily to precise the question.
If I understand you correctly, you're asking for an $L$-theory $T$, presumably extending the theory of fields, such that every model of $T$ is of characteristic $0$ and hence contains $\mathbb{Q}$ (this is easy) and contains a transcendental element (this is harder).
There are such theories, but analyzing them requires a lot of algebraic work. As a concrete example, let $R$ be the real closure of $\mathbb{Q}$ (its elements are the real numbers which are algebraic over $\mathbb{Q}$). Now consider $R(t)$, the field obtained by adjoining one transcendental $t$. It's a fact that $R$ is definable in $R(t)$. That is, there is a first-order formula $\varphi_R(x)$ in the language of fields such that for all $a\in R(t)$, $R(t)\models\varphi_R(a)$ if and only if $a\in R$.
I can even write down the formula explicitly: $\exists y\, 1+x^4 = y^2$. For a proof that this works, see Proposition 3.3 in this book, noting that every real closed field is Pythagorean (every sum of two squares is non-negative, hence is a square) of characteristic $0$.
Now let $T = \text{Th}(R(t))$. $T$ expresses the following:
So any model $K\models T$ has a subfield $\varphi_R(K)$ which is real closed, and hence contains $\mathbb{Q}$, but $K$ also contains elements which are not algebraic over $\varphi_R(K)$, and hence are not algebraic over $\mathbb{Q}$.