Plectis

Comparator case study

Comparator #1049 fixture

Rendered from docs/COMPARATOR_1049_NUMERICAL_HEIGHT.md in the public Lean repository. View source

Erdős #1049: one numerical certificate, exactly scoped

The question

For which rational bases is the relevant Lambert-type series irrational? The base 3/2 is the main unresolved target in this programme.

What is checked here

Lean proves the concrete inequality

log(2) / log(7) < 1/2 - 1/pi^2.

Comparator then checks a tiny proof transport from the public source theorem to a separately declared Mathlib-only statement. A deliberately weakened nearby statement must be rejected.

This is useful because the inequality is the numerical input used at the base 7/2 in an external theorem of Bundschuh and Väänänen. It cleanly separates a kernel-checked arithmetic certificate from the literature theorem that consumes it.

What is not checked here

Lean does not formalise the full Bundschuh-Väänänen theorem, all of its hypotheses, or its irrationality conclusion. This packet does not advance the open base 3/2 case.

Who did what

AI agents helped draft the Lean proof, the statement firewall, and this exposition. Will Cook directed and reviewed the public release and remains responsible for its claims. The Lean kernel checks the source proposition. Comparator checks the transport to the separately declared statement. The Bundschuh-Väänänen theorem remains a cited external result.

Proof handle

The correct badge is: Comparator-checked numerical certificate. It is not an independently verified irrationality theorem.