Vaught's conjecture is a famous open problem in mathematical logic from 1961. It asks whether a complete logical theory (a consistent set of axioms written in a countable language) can have exactly some uncountable but not maximum number of distinct infinite models. More specifically, it predicts that any such theory must have either countably many or the maximum possible number of countably infinite models, with nothing in between. Despite decades of effort, the full conjecture remains unsolved, though it has been proven for certain restricted classes of theories.
This paper introduces a new refined version of the conjecture that connects two existing lines of research: Martin's conjecture, which concerns the complexity of functions on real numbers under a framework called Turing reducibility, and the omega-Vaught's conjecture, which studies how complicated a description (called a Scott sentence) you need to uniquely pin down a mathematical structure. The authors combine these ideas into what they call the "type omega-Vaught's conjecture," which predicts not just how many models a theory has but also how logically complex those models are in a precise, calibrated sense.
The main result is a proof of this new conjecture for a well-behaved class of theories called omega-stable theories. Along the way, the authors carefully re-examine two classical proofs from the 1980s and 1990s, extracting sharper information from them than was previously recorded. A key concrete improvement is that when an omega-stable theory has only countably many countable models, each such model can be described by a Scott sentence of much lower logical complexity than previously shown, requiring roughly omega plus 5 quantifier alternations rather than omega plus omega. This kind of complexity bound matters because it gives a tighter picture of just how hard it is to distinguish mathematical structures from one another using logical formulas.