Is the following (second-order) formula schema provable in ATR$_0$?
Let $\varphi$ be an arithmetical formula satisfying
- For all $x, y\in \mathbb{R}$, we have that $x=_\mathbb{R}y$ implies $\varphi(x)\leftrightarrow \varphi(y)$.
- For all $x\in \mathbb{R}$, if $\varphi(x)$, then $(\exists N\in \mathbb{N})(\forall y\in B(x, \frac{1}{2^N}))\varphi(y)$.
Then there are sequences of reals $(a_n)_{n \in \mathbb{N}}$ and $(b_n)_{n \in >\mathbb{N}}$ such that for all $x\in \mathbb{R}$, we have $\varphi(x)\leftrightarrow (\exists n\in \mathbb{N})(a_n<x<b_n)$.
The idea behind the previous schema is that $\varphi(x)$ defines an open set of reals via 1) and 2), and therefore has a code in the sense of second-order reverse mathematics.